COST Action ES1002 Weather Intelligence for Renewable Energies (WIRE)

CURRENT STATE Report August 2012

Contacts: Dr. Alain Heimo Chair COST Action ES1002 “WIRE” Email: [email protected] or List of authors in Chapter 8

Table of Content

1 Management summary ...... 7

2 Political, economical and technical framework ...... 10

3 Research and Development: the European approaches ...... 17

3.1 Wind Energy ...... 18

3.2 Solar Energy ...... 39

3.3 Grid management ...... 68

3.4 Wave Energy ...... 80

4 National activities ...... 82

4.1 AUSTRIA ...... 83

4.2 AUSTRALIA ...... 85

4.3 BELGIUM ...... 88

4.4 BOSNIA AND HERZEGOVINA ...... 90

4.5 BULGARIA ...... 93

4.6 CROATIA ...... 97

4.7 CZECH REPUBLIC ...... 100

4.8 DENMARK ...... 104

4.9 FINLAND ...... 109

4.10 FRANCE ...... 111

4.11 GERMANY ...... 116

4.12 GREECE ...... 121

4.13 HUNGARY ...... 124

4.14 ICELAND ...... 127

4.15 ISRAEL...... 129

4.16 ITALY ...... 132

4.17 THE NETHERLANDS ...... 136

4.18 NORWAY ...... 139

4.19 POLAND ...... 141

4.20 ROMANIA ...... 145

4.21 SPAIN ...... 148

4.22 SWITZERLAND ...... 153

4.23 TURKEY ...... 156

5 Conclusions / Recommendations ...... 159

6 Appendix A ...... 161

7 Appendix B ...... 163

8 List of authors ...... 164

Table of Figures

Picture 1: Instruments allowing the inference of the diffuse and direct component of solar radiation separately; a) rotating shadow-band instrument, b) instrument with shading pattern. (courtesy N. Geuder) ...... 14

Figure 1: Typical errors introduced by the NWP. Source: Möhrlen [18]...... 21

Figure 2: The development of the forecast error during the last years in the E.On Netz area. The numbers in square brackets are references from Lange et al. [19]...... 22

Figure 3: The error comes from the NWP. The figure shows the difference in degree of explanation between Sipreolico run with HIRLAM input (from an older version of the Spanish HIRLAM) and Sipreolico run with on-site wind speed input. Source: Sanchez et al. [26]...... 24

Figure 4: Very high resolution model domains from left to right: AROME (Meteo France, 41 vertical layers), COSMO-DE (DWD, 50 vertical layers), UM-4km (grey shaded area, UK Met Office, 70 vertical layers)...... 25

Figure 5: Spatial cross-correlation of prediction deviations for various prediction times based on German data for the years 1996–1999. For comparison the cross-correlation coefficients of the prediction (36 h) are also shown. All cross-correlation coefficients have been averaged over 25km bins. The figure is provided by M. Lange, energy & meteo systems GmbH...... 29

Figure 6: Forecast error standard deviation ratio versus region size quantified by the region diameter. The horizontal line gives the expected error reduction for an area the size of Germany. The figure is provided by M. Lange, energy & meteo systems GmbH...... 30

Figure 7: Typical model chain for PV power prediction ...... 43

Figure 8: Short-term forecasting scheme using cloud index images...... 45

Figure 9: Nested domains of the WRF model used in [40]. The outermost domain (green), middle domain (red), and the inner study area (blue) have spatial resolutions of 27 km×27 km, 9 km×9 km, and 3 km×3 km, respectively...... 47

Figure 10: Conversion steps for estimating global irradiance on tilted surfaces from global horizontal irradiance. The components framed in red are empirical models introducing further sources of uncertainty...... 50

Figure 11: Example of normalized MPP efficiency as a function of irradiance on the module plane. Blue: MPP efficiency for Tmodule = 25°C. Red:

MPP efficiency for Tambient = 15 °C. Green: AC efficiency for Tambient = 15 °C. The curves are normalized to the efficiency at standard test −2 conditions (STCs). STCs are defined by Tmodule = 25 °C, Irradiance = 1000 Wm , and the standard spectrum of air mass AM1.5...... 51

Figure 12: Left: Time series of predicted and measured global irradiance for the period from 29 April 2007 to 6 May 2007. Right: Scatter plot of predicted vs. measured global irradiance...... 53

Figure 13: Relative RMSE of ECMWF-based forecasts and of reference models depending on the forecast horizon...... 54

Figure 14: Bias (left) and standard deviation of the error (right) of the irradiance forecast in Wm-2 as a function of the cosine of the solar zenith angle (cos ΘZ) and the clear-sky index kt*...... 55

Figure 15: Forecast of global irradiance Iglob with confidence intervals of an uncertainty level of 95% compared with measured irradiance for six days in May 2007 for a single site (left) and for the average of two hundred measurement stations in Germany (right)...... 56

Figure 16: RMSE of the five forecasting approaches and persistence for all three German stations for the first three forecast days (left) and for the single stations for the first forecast day only (right). The average hourly global irradiance for all stations including only daylight hours was 227 Wm−2...... 57

Figure 17: Absolute (left) and relative (right) global irradiance forecast errors: RMSE (solid line with circles) and bias (dashed lines) of five different forecasting approaches and persistence in dependence on the month for the first forecast day using data from three German sites...... 58

Figure 18: RMSE for the first, second, and third forecast day for stations in Spain. The average hourly global irradiance including only daylight hours was 391 Wm−2...... 58

Figure 19: Relative RMSE of global irradiance forecast for single sites (left) and regional average values (right) based on motion vectors from satellite images (orange) compared with global model forecasts ECMWF-OL (dark blue), satellite-based irradiance values (light blue), and persistence (red)...... 59

Figure 20: Left: Correlation coefficient of forecast errors of two stations depending on the distance between both. Blue dots: observations, red dots: exponential fit function. Right: Error reduction factor f = RMSEensemble/RMSEsingle for regions with increasing size...... 61

Figure 21: Definition of Meteorological Icing and Instrumental Icing...... 77

Figure 22: Electric power generation and power consumption – public power grid from 1990 to 2009 [1]. Source: Energie-Control GmbH ...... 83

Figure 23: Growth of the accepted photovoltaic installations in the Austrian states. Source: Energie-Control GmbH ...... 84

Figure 24: The renewable resource in Denmark, according to the Climate Commission 2010 ...... 104

Figure 25: The Danish wind resource map. Source: Risø and EMD. See http://www.emd.dk/windres...... 105

Figure 26: The share of wind power in the grid (orange line). The green and blue bars denote installed capacity onshore and offshore, respectively. Source: Energistatistik 2010, Energistyrelsen...... 106

Figure 27: The changing structure of the Danish electricity production between 1985 and 2009. Source: Energistyrelsen...... 107

Figure 28: German annual renewable electricity generation in GWh from hydro (blue), wind (turquoise), bioenergy (green), and photovoltaics (yellow). 1990–2011 [4]...... 117

Figure 29: German annual electricity generation from wind turbines (columns in GWh) and installed wind power capacity (solid curve in MW). 1990–2011 [4]...... 117

Figure 30: German annual electricity generation from photovoltaics (columns in GWh) and installed PV capacity (solid curve in MW). 1990–2011 [4]...... 118

Figure 31: German annual electricity generation from biomass in GWh. 1990–2011 [4]...... 118

Figure 32: German annual electricity generation from hydro power plants (columns in GWh) and installed hydro power capacity (solid curve in MW). 1990–2011 [4]...... 119

Figure 33: Average wind speed at 75 m.a.g. over Hungary (Source: Hungarian Meteorological Service, 2005) ...... 125

Figure 34: A real distribution of yearly total of sunshine duration in Hungary (Source: Hungarian Meteorological Service, 2010)...... 125

Figure 35: Areal distribution of yearly totals of total global radiation in Hungary (Source: Hungarian Meteorological Service, 2010) ...... 126

Figure 36: Primary energy use in Iceland 1940-2010 [1]...... 127

Figure 37: Share of renewable energy from national energy sources in the Netherlands (in % of the total electricity consumption in NL). Source :CBS Stateline NL ...... 137

Figure 38: (Left) Installed capacity by technologies (in %) of the Spanish electric. (Right) Contribution to the electricity demand of different technologies (in %). Source: Spanish Grid operator www.ree.es...... 148

Figure 39: Evolution of the electric installed capacity in Spain, for different technologies, along the period2006-2010. Source: Spanish Grid operator www.ree.es...... 148

Figure 40: Percentage of the world wind energy installed capacity by 2010. Source: Spanish Wind Energy producers association (http://www.aeeolica.org/) ...... 149

Figure 41: Evolution of the wind energy production (green) and share (blue) of the wind power in Spain along the last decade. Source: Spanish Wind Energy producers association (http://www.aeeolica.org/)...... 149

Figure 42: Evolution of the solar PV power installed capacity (yellow) in Spain along the last decade. Source: Spanish PV Energy producers association (http://www.asif.org)...... 150

Figure 43: Contribution of the PV power to the electricity demand as a function of the year month, for the year 2009 (red) and 2010 (green). Source: Spanish PV Energy producers association (http://www.asif.org)...... 150

Figure 44: Evolution of the solar CSP power installed capacity (blue) and their contribution to the electricity demand (yellow) in Spain along the last years. Source: Spanish CSP Energy producers association (http://www.protermosolar.com)...... 151

Table of Tables

Table 1: Hydro and other renewable energy production 2010 – 2035 (source: IEA Observ’er 2011) ...... 11

Table 2: Relative bias and relative RMSE of different regional forecasting approaches for the German control areas of 50 Hertz and TenneT (for all 24 h of the day)...... 62

Table 3: Relative bias and relative RMSE of different forecasting approaches for single PV systems within the control area of 50 Hertz (for all 24 h of the day)...... 62

Table 4: Installed capacity and production in 2011 of wind, solar and hydro electrical powers for some member countries. The last 2 columns indicate the share of all kinds of renewable energies to the national productions of electricity in absolute and relative units...... 82

Table 5: Croatia’s electricity balance in 2009/2010...... 97

Table 6: Installed capacities for heat and electricity generation from RES in Croatia for 2010 (preliminary data)...... 98

Table 7: Czech Republic renewable energy potential ...... 100

Table 8: The recent status and near future estimation of power from renewable energy sources (source: www.ypeka.gr) ...... 121

Table 9: The estimated achievable renewable energy potential of Hungary (Source: Hungarian Academy of Sciences, Renewable Energy Subcommittee, 2009) ...... 124

Table 10: Predicted energy production in Israel for the period from 2014 – 2020 by available renewable energy technologies [7]. PV stands for photo-voltaic solar technology, TS stands for thermo-solar technology...... 129

Table 11: Net electricity production in Italy (Source: Terna and GSE) ...... 132

Table 12: Net Power in Italy (Source: Terna and GSE) ...... 133

Table 13: Poland – renewable sources energy – march 2011 [1] ...... 141

Table 14: The goals set up in the document "National Plan of Action on Renewable Sources of Energy " [3] ...... 142

Table 15: Estimation of energy from renewable sources [2] ...... 142

Table 16: Energy potential of the Romanian renewable energy sources ...... 145

Table 17: Estimated development of renewable electricity production in Switzerland,3...... 153

Table 18: Turkey’s renewable energy potential ...... 156

1 MANAGEMENT SUMMARY

The world population is constantly increasing and the world electricity consumption will presumably double by 2050 with potential dramatic effects on our climate. It is expected that worldwide primary energy demand will increase by 45%, and demand for electricity will increase by 80% between 2006 and 2030 [1] Consequently, without decisive action, energy-related greenhouse gas (GHG) emissions will more than double by 2050, and increased oil demand will intensify concerns over the security of supply. There are different paths toward stabilizing GHG concentrations, but a key issue in all of them is the replacement of fossil fuels by renewable energy sources.

The EU's dependence on imports of fossil fuels (natural gas, coal and crude oil) from non-EU countries, as a share of total primary energy consumption, rose from 50.8 % in 2000 to 58.2 % in 2009 [2] . In additions, baseline scenarios show a rising dependence on imports for most fossil fuels, although this is particularly relevant for gas, with forecasted imports (as a percentage of primary energy consumption) rising from around 58.2 % in 2009 to up to 84 % by 2030. In order to correct this situation, and considering that many countries have decided to lessen their dependence on nuclear energy, the European Union has adopted the goal of having 20% of its electricity supply from renewable energy sources by 2020, along with a commitment to achieve at least a 20% reduction of greenhouse gases by 2020, compared to 1990 (European Directives 2009/28/EC and 2009/29/EC).

Wind and solar power are presently considered as the sources of renewable energy with the best chance to compete with fossil-fuel energy production in the near future and the configuration of the penetration of different sources of electricity is rapidly evolving: for example, in March 2011 the Spanish wind farms have produced 4738 GWh of electricity, covering 21% of the demand or the consumption of 3 million homes: this monthly energy production would meet the needs of a smaller country such as Portugal. Such an impressive example shows that the penetration of renewable energies in Europe is on the right track.

However, the optimum integration of electricity produced by future wind turbines and solar power plants demands an accurate wind and solar energy potential availability evaluation and forecast. What happens when these conditions are not met? As an example, a 10% of uncertainties in the estimates of mean wind might lead to a 30% error in power production. Wind and solar energy potential evaluation and forecasting as well as electrical grid management studies aim to answer such questions and have the goal to help developers of renewable energy power plants to decide where to install and how to operate them most efficiently and to help the grid operators to manage this per definition intermittent production input more efficiently.

Long-term averages based mainly on historical measurement data are usually used for the resource assessments of the sites where wind farms or solar power plants will be installed. Once this is done, forecasting tools are applied for short-term information - i.e. 1- to 72 hours in advance depending on the usage - on the production. Indeed, for the operational management of electrical grids, integrating different power sources and dealing with the highly spatially distributed locations of the power plants together with the intermittency, weather dependent production becomes a very important aspect and determines if the production will remained balanced with the demand. For example, Denmark has already a high penetration of wind energy production: a small change in wind speed may result in a considerable change in the power production.

Different time scales for renewable energy production forecasting have to be considered. For very short-term prediction (from 30 minutes to 3 hours), persistence forecasting is presently used in the case of wind energy: it is based on the simple assumption that the wind speed will usually not change dramatically in the very short-term. The situation is different - and more difficult - for solar energy where extremely rapid changes in the local cloud cover may induce dramatic changes in the output of the individual power plant.

To improve the very short-term forecasts, the present trend is to use Numerical Weather Prediction NWP models, combined with post-processing techniques such as downscaling and the on-line assimilation of in-situ and regional

ground-based or remote sensing measurements, as demonstrated by the CNMET project in Switzerland [3]. The situation is more difficult for the prediction of solar energy where improvements of the cloud-tracking algorithms (a combination of satellite and ground-based measurements) are urgently needed. Furthermore, these very-short term predictions are much more difficult in complex terrain where the topography may dramatically decrease the accuracy of the prediction (turbulence effects, local wind conditions, etc.). Finally, harsh weather conditions (e.g. icing) is a further weather dependent source of potential failures for energy production and distribution.

For short-term forecast in the time horizon 3 to 72 hours, persistence should not be applied anymore. Extensive uses of numerical weather predictions with appropriate post-processing algorithms are usually implemented. Here again, the wind energy production forecasting systems are more advanced than for solar energy, because of the improvements of the capability of NWP models to predict the wind conditions days ahead with a reasonable accuracy in contrast to cloud cover prediction and aerosol loads which are particularly important for the solar energy production.

From the electrical grid point of view, the situation is different: the penetration of renewable energies implies more “intelligence” in order to manage their integration and to guarantee continuously the equilibrium between production and demand. Electrical grids were built in Europe more than 50 years ago from a strictly national point of view. Existing electrical grids are very centralized, transferring the power between big power plants towards the end users; however, the number of relatively small decentralized production units will increase dramatically.

What is today needed is an “exportable” approach which would allow increasing electricity transfers amongst grids at different levels from local to national to European. Meteorological conditions in Europe are such that the wind is likely to blow or the sun is likely to shine at some place in Europe: in order to increase the penetration of renewable energies, it is mandatory to consider the electricity exchanges on a more extensive scale: these will be efficiently managed only by introducing "intelligent" technologies such as “smart grids”.

The challenge for the electrical grid operators is to synchronize at every moment the energy production with the demand. This equilibrium is constantly changing with the fluctuation of the demand and it is further jeopardized by the increasing penetration of renewable energy sources such as solar and wind whose variability induce significant energy fluctuations on the grid. The difficulty to accurately forecast (intermittent) renewable energy production is an increasing challenge for the Transmission System Operators TSO which have to cope with the dangerous risks of grid instabilities: already nowadays wind or solar power plants may have to be disconnected from the grid in case of low demand unless included in a smart grid system where the excess of energy may be planned for alternative uses or stored. Accurate power forecasting, efficient and intelligent grid management and increased flexible storage capacity are mandatory for the efficient development of the future energy policies in Europe and worldwide, not to mention the benefits in terms of climate change when considering the greenhouse emissions of the respective energy productions: for example, energy produced with coal emits about 85 times more CO2 than wind energy, petrol 70 times and natural gas 40 times.

Intelligent management systems ("smart grids") will aim at adapting in real time and efficiently the energy production to the fluctuating demand. Storage will then be able to combine centralized and decentralized (renewable energies) production systems. For this purpose, short-term weather - and production - forecasts will play a major role when considering the whole of Europe.

Another demanding issue is the potential of power line to transport energy as function of the Dynamic Line Rating DLR that has to be considered in terms of efficiency. The fact that renewable energies will be often produced in remote and decentralized sites implies that the electrical grid will be extended with power lines often running through areas with different(often harsh) weather conditions. DLR has to be taken into account in order to design these new installations in an optimal way and to operate them efficiently by taking into account their environmental conditions (e.g. icing). Standard transmission and distribution lines have however an inherent reserve capacity. The question of DLR may nevertheless be raised for “bottlenecks” in the system, or when the reserve capacity is limited

during peak load hours. Such peak loads may occur during cold winter days in northern Europe when there is a high power demand for house heating or during hot summer days in southern Europe when there is likewise a high demand for air conditioning.

Finally, financial aspects have also to be considered. For independent wind-farm or solar power plant operators selling their electricity directly on the market, inaccurate production forecasts make the difference between large profits and large fines for non-compliance. For grid operators, accurate forecasting will result in less fossil-fuel based energy production kept "burning" and a better efficiency in the use of storage capacity.

The COST Action ES1002 “Weather Intelligence for Renewable Energies WIRE” was launched in November 2010 to promote the short time forecasts of energy production for wind and solar energy. Its goals are 3-fold: evaluate the accuracy of existing forecast systems (including post processing algorithms) by validating their results with in-situ measurements performed mainly at power plant sites (Working Group 1), promote the use of ground-based standard and remote sensing measurements together with satellite-borne information and analyze the potential to increase the quality of the short-term forecast with such systems (Working Group 2) and finally strengthen the collaboration between end-users (power plant operators and TSOs) and modelers in order to best characterize their needs and requirements (Working Group 3) based on the modeling results.

Two major strategies are applied: first, solar and wind energies are considered in a single approach reflecting the worldwide challenges set by a high penetration of renewable energies from the electrical grid point of view. Second, a European-wide approach is promoted reflecting the fact that the most efficient way to manage the growing share of wind and solar energies will be to consider Europe as a whole.

The present document reflects these goals: a first part is dedicated to the general aspects linked to the increasing penetration of renewable energies. A second part deals with the latest achievements in the fields of production forecasts and integration in the existing electrical grids. A third chapter presents the situation in the different member countries. Finally, open questions and recommendations are presented which reflect the goals and planed activities of the present COST Action ES1002 “WIRE”.

References to Management Summary:

[1] IEA, 2009 World energy outlook. International Energy Agency, OECD publication service, OECD, Paris. [2] EEA, 2012. Net Energy Import Dependency (ENER 012) - Assessment published Apr 2012 European Environmental Agency Report EEA Report No 6/2012 [3] Calpini, B., Ruffieux, D., Bettems, J.-M., Hug, C., Huguenin, P., Isaak, H.-P., Kaufmann, P., Maier, O., and Steiner, P.: Ground-based remote sensing profiling and numerical weather prediction model to manage nuclear power plants meteorological surveillance in Switzerland, Atmos. Meas. Tech., 4, 1617-1625, doi:10.5194/amt- 4-1617-2011, 2011

2 POLITICAL, ECONOMICAL AND TECHNICAL FRAMEWORK1

Due to the increasing world population and economic growth, global demand for energy is increasing rapidly: the world population has exceeded 7 billion at the end of 2011 and the global economy grew 3.3% per year over the past 30 years. Hence energy demand increased 3.6% in the world

Conventional energy has been used from ancient times, their sources consisting primarily of wood, coal, natural gas and oil from decaying plants and animals over hundreds of thousands to millions of years. More recently, the Industrial Revolution has been made possible by the use of conventional energy sources.

Carbon-based natural gas, oil and coal store carbon as potential energy and release it when burned. Currently, coal, oil, and natural gas supply nearly 88 percent of the world’s energy needs; for example, oil provides about 41% of the world’s total energy supply.

Coal is one of the significant conventional energy sources. According to Index Mundi of World Coal Consumption, the world coal consumption exceeded 8 billion tons in comparison with 4.9 billion tons in 1990. The major part is consumed in China with 3.5 billion tons followed by USA with 1 billion tons. Following IEA estimations, coal consumption is expected to increase between 25% (minimum scenario) and 65% (maximum scenario) till 2035 [1].

A comprehensive assessment of global fossil-fuel subsidies has found that governments are spending $500 billion annually on policies that undermine energy security and worsen the environment [2]. Generally speaking, economic risk of relying on imported energy has grown as oil prices have become turbulent. Hence, the rising prices have had a negative impact on poor countries where most of the fuel and many of the technologies are imported. The negative effect of oil price on the macro-economy is significant, and should be used to build the business case to invest in alternative energy carriers [3].

From the climate change point of view, the burning of fossil fuels leads to increased greenhouse gas emissions and other environmental damages. According to the OECD Environmental Outlook to 2050, global greenhouse gas emissions continue to increase: currently, atmospheric concentration of CO2 has reached 392 ppm with an increase of about 2 ppm per year. In 2050, this concentration will be over the level of 450 ppm required to have at least a 50% chance to stabilize the climate within a 2°C global average temperature increase. This is mainly due to global energy- related CO2 emissions which reached 32 Gt in 2010 despite the recent economic crisis [4]. This level is all-time peak of CO2 emissions and 45% higher than in 1990 and indicates that the energy policies, energy supply and consumption have major effect on occurring of greenhouse gas emissions because the energy-related CO2 emissions represent the majority of global greenhouse gases. In recent COP17/CMP7 meeting in Durban, countries agreed on an advanced framework for the reporting of emission reductions for both developing and developed countries taking into consideration the common but differentiated responsibilities of different countries [5].

With the start of the Kyoto Protocol, the world entered a new developing model with new targets covering sustainable living with renewable energy technologies. At this point, the climate change represents a fundamental challenge to the global sustainable development. To reduce these concerns, global communities are trying to find and implement different energy saving strategies and technologies in parallel to alternative sources of energy for different sectors that rely on energy produced from different sources. In this regard, wind and solar energy development will play a significant role to meet future energy demands and contribute to the reduction of the environmental pollution.

1 Lead author: Selahattin Incecik

Indeed, there is a significant progress on wind power worldwide. At the end of 2011, the worldwide installed capacity of wind power has reached 240 GW to generate more than 430 TWh annually which is about 2.5% of world electricity consumption. There are only a few countries that made big progresses on wind power: USA and China became the largest wind energy providers worldwide. During the last 5 years, China’s total capacity significantly increased and wind power is considered as a key growth component of the country's economy: at the end of 2011, the wind power capacity in China accounted for 63 GW while only 2.6 GW in 2006. The other largest wind energy provider country is the USA where the cumulative installed capacity is 47 GW and wind power accounted for 2.3% of the electricity generated in 2011. In the EU, Germany has largest installed capacity with 29 GW which represents about 6.5% of the nationally generated electricity. Finally, India is rapidly increasing its wind energy installed capacity with 16GW (5th place in the world).

From a broader point of view, and apart from wind energy, the recent IPCC Special Report on Renewable Energy Sources and Climate Change Mitigation (2011) indicates that deployment of renewable energy has increased rapidly in recent years. Renewable energy installed capacity continued to grow rapidly in 2009 compared to the cumulative installed capacity from the previous year, including hydropower (3%; 31 GW added), grid-connected photo voltaic (53%; 7.5GW added), geothermal power (4%; 0.4 GW), and solar hot water heating (21%; 31 GW added).

Looking at the future, the following Table 1 presents the latest IEA projections of the world renewable energy productions for 2035 compared to 2010:

Table 1: Hydro and other renewable energy production 2010 – 2035 (source: IEA Observ’er 2011)

2010 2035

4158 TWh2 8232 TWh Total 19.6% of total electricity 23.4% of total electricity production production

[TWh] [%] [TWh] [%] Hydro power 3448 82.9 5260 68.3 Biomass 263 6.3 773 9.4 Geothermal 68 1.6 186 2.3 Wind 345 8.3 1462 17.7 Solar 33 0.8 191 2.3

References to §2:

[1] IEA, 2011, World Energy Outlook 2011, Paris, OECD. [2] REN21, 2005 Renewables 2005 Global Status Report, Renewable Energy Policy Network by Worldwatch Institute [3] Owen, N.A., Inderwildi, O.R. , King, D.A., 2010, The status of conventional oil reserves – hype or cause of concern? Energy Policy, 38, 4743-4749. [4] OECD, 2011, Joint report by IEA, OPEC, OECD and World Bank on fossil-fuel and other energy subsidies: An update of the G20 Pittsburgh and Toronto Commitments http://www.oecd.org/env/49090716.pdf

2 1KWh = 3.6 106 J

[5] COP17, 2011, United Nations Climate Change Conference, Durban, South Africa, 28 Nov - 9 Dec 2011, Powering Climate Solutions

2.1.1 CLIMATE CHANGE AND VARIABILITY ON WIND AND SOLAR RESOURCES3

The surface winds are mainly driven by large scale circulation. However, several local features such as the surface roughness, orography and thermal contrasts modify the spatial and temporal features of the surface winds. Regional variability of wind speeds is also controlled by the interaction of large scale dynamics and orography.

In the last decades, it has been recognized that climate change may induce wind speed trends. There are several papers ([1], [2], [3]) on the decreasing surface wind speeds in some regions of the globe such as USA, Australia, China and some parts of Europe. However, no clear trends can be recognized in Scandinavia. Climate change may impact the surface winds by changing the geographic distribution and / or the annual and inter-annual variability of the wind resource ([4]).

The Intergovernmental Panel on Climate Change (IPCC) states that there is evidence for long-term changes in the large-scale atmospheric circulation, such as a pole-ward shift and strengthening of the westerly winds [5]. It is also expected that these observed changes will continue. According to recent reviews of the historical trends of jet streams, they will rise in altitude and move pole-ward in both hemispheres ([6], [7]). Furthermore, the northern hemisphere jet is weakening.

Consequently, the reductions in wind speeds would decrease the available wind energy. Furthermore, climate change will affect seasonal cloud cover and impact the available solar resource on the Earth surface. Therefore, investigating the wind speed and cloud cover changes can increase our understanding of the climate change relationships [8], [9], [10], [11].

References to §2.1.1

[1] Pryor SC, Schoof JT, Barthelmie RJ. 2006, Winds of Change? Projections of nearsurface winds under climate change scenarios. Geophysical Research Letters, 2006;33(L11702). doi: 10.1029/2006GL026000. [2] Guo, H., M.Xu, Q.Hu, 2011, changes in near-surface wind speed in China: 1969-2005. Int.J.Climatology, 31, 349-358. [3] Najac J, C.Lac, L.Terray, 2011, Impact of climate change on surface winds in France, using a statistical – dynamical downscaling method with mesoscale modeling, Int. J. Climatology, 31, 415-430. [4] S.C. Pryor , R.J. Barthelmie, 2010, Climate change impacts on wind energy: A review, Renewable and Sustainable Energy Reviews 14, 430–437 [5] Eichelberger, S. J. and D. L. Hartmann, 2007, Zonal jet structure and the leading mode of variability. J. Climate, 20, 5149–5163, doi:10.1175/JCLI4279.1. [6] Archer C.L and K.Caldera, 2008, Historical trends in the jet streams, Geophysical Research Letters, 35, L08803,doi: 10.1029/2008GL033614 [7] Kidson J. and E.P.Gerber, 2010, Intermodel variability of the poleward shift of the austral jet stream in the CMIP3 integrations linked to biases in 20th century climatology, Geophysical Research Letters,37. L09708, 5 PP., 2010 doi:10.1029/2010GL042873

3 Lead author: Selahattin Incecik, David Pozo Vasquez

[8] Schaeffer,R, A. S.Szklo, A. F. P. de Lucena, B. S. M. Cesar Borba, L. P. P.Nogueira, F. P. Fleming, A. Troccoli, M. Harrison, M. S. Boulahya, 2012, Energy sector vulnerability to climate change: A review, Energy, 38, 1-12. [9] Matuszko, 2011, Influence of the extent and genera of cloud cover on solar radiation intensity, Int J Climatology, DOI: 10.1002/joc.2432. [10] Cutforth HW, Judiesch D. 2007, Long-term changes to incoming solar energy on the Canadian Prairie. Agricultural and Forest Meteorology, 145:167-75. [11] Bartók B. 2010, Changes in solar energy availability for south-eastern Europe with respect to global warming. Physics and Chemistry of the Earth,35,63-69

2.1.2 THE WORLD METEOROLOGICAL ORGANIZATION APPROACH4

Dedicated meteorological products are designed for the energy sector, including specific measurement techniques, and the need for a corresponding “regulatory” framework is emerging (e.g., standard operating procedures, certification, traceability, etc.). For meteorology, a general framework for observation techniques is defined by the Commission for Instruments and Methods of Observations (CIMO) in its Guide for Meteorological Measurements (“CIMO Guide”). This document defines accuracy requirements for different type of measurements, standard operating procedures and certification guidelines for sensors and measurement systems. Proper characterization of measurement techniques tuned to the energy sector requires several steps, the first one being the definition of the parameter of interest. The Guide to the expression of uncertainty in measurements [1] describes this as specification of the measurand. For example, the diffuse and direct components of solar radiation are parameters of interest for the solar energy sector, rather than the horizontal global irradiance more commonly used in meteorology. Definition of the parameter of interest should be complete, including for instance geometrical or spectral specifications (e.g., collimation, spectral range etc.). For wind energy, parameters of interest include all measurements that are relevant for estimating the mean wind at the hub height of the wind turbine i.e. direct measurable such as temperature or derivate such as the atmospheric stability that controls the shape of the vertical wind profile i.e. the variation of wind with height. The estimate of atmospheric turbulence indices is needed to characterize the location in order to choose the most suitable turbine class for the site. Furthermore, atmospheric turbulence is also critical to study the interaction wakes and turbulence and to model the wind turbine wakes in different atmospheric stability conditions in order to forecast the power lost due to the upwind wakes. Once the parameters of interest are identified, the observational requirements of the energy sector should be clarified. This includes the measurement frequency, the required accuracy, the minimum level of data availability, and eventually other characteristics specific of the domain of interest (e.g., local representativity, spatial extent of the measurement, etc.). Such observational requirements must be defined considering the purpose that the measurement should serve in applications for the energy sector. This allows combining end-user requirements and uncertainties of the inference process for defining the measurement requirements. They may of course be different for every application, but general requirements can often be defined for groups of applications. Procedures are then defined for evaluating instrument performance and standard operating procedures are elaborated for reaching the specified requirements. This eventually leads to defining standards (e.g., ISO standards). In meteorology, such tasks are devoted to bodies within the World Meteorological Organization (WMO), typically the CIMO. In general these are carried out jointly with other international bodies such as the Bureau international des poids et mesures (BIPM). In the case of meteorological measurement for the energy sector, it is judicious to also involve the International Energy Agency (IEA).

2.1.2.1 SOLAR ENERGY

4 Lead author: Bertrand Calpini and Laurent Vuilleumier

Different geometries are used for the solar energy collection devices such as solar concentrators, photovoltaic panels with various orientations or solar thermal panels. Thus, as mentioned above, the information about the overall solar energy flux over a horizontal surface (global horizontal irradiance) is not sufficient for assessing the solar energy input onto the collection device. It is better to know separately the solar direct and diffuse radiation components since one can reconstruct the radiance distribution with this information and limited assumptions on the distribution of the diffuse radiance. On the other hand, accuracy requirements are not as strict as for other meteorological applications, typically climate change monitoring. Such accuracy requirements are not yet completely clarified. But activities ongoing within the framework of COST WIRE will contribute to their better definition.

Preliminary enquiries indicate that desirable characteristics or performance from instruments measuring solar radiation for the solar energy sector include: a) ability to distinguish the direct and diffuse component of solar irradiance; b) time resolution 10–15 min; c) RMSE ≤ 5% (at 10min resolution); d) data availability > 90%. In addition, a measurement frequency on the order of min-1 (with sampling frequency of sec-1 and statistics such as mean, standard deviation, minimum and maximum during the integration period) allows obtaining information on ramps. Within the framework of the COST Action ES1002 WIRE, a performance evaluation of instruments allowing inferring the solar energy input on different type of surfaces is foreseen. Part of this activity will serve to define more precisely requirements for such instruments. Besides allowing the inference of the diffuse and direct component of solar (shortwave) radiation separately, instruments of interest for the solar energy sector should operate in a robust and cost effective way without the use of sun trackers and maintenance-intensive sensors: They are usually deployed in the field for continuous operation with limited maintenance. Two kinds of such instruments are on the market. The first type uses a rotating shadow-band that alternately shades and then exposes the entrance aperture of the instrument (see Fig. 1a). Such measurement cycles allow estimating the global solar irradiance and the diffuse irradiance component with proper algorithms. The direct-normal component is then inferred from the difference of the two measurements. The other type of instrument uses an elaborate computer-generated shading pattern and an array of thermopile sensors (see Fig. 1b). These instruments are designed so that for almost any position of the sun in the sky, some sensors are exposed to the direct sun and some are in the shade. This also allows inferring the global solar irradiance and its diffuse component, and subsequently the direct normal irradiance. For solar radiation measurements, some existing standards are ISO 9060, defined by the Technical Committee ISO/TC 180 [2]. These standards mainly concerns measurements for meteorological or climate research purpose. Standards for instruments tuned for the solar energy sector (as described above) are still missing.

2.1.2.1.1 PERFORMANCE EVALUATION FOR SOLAR ENERGY5

The necessity of evaluating the performance of the above-mentioned instruments was pointed out within the COST Action ES1002 WIRE. The Baseline Surface Radiation Network (BSRN) station of Payerne has been proposed to act as a WMO/CIMO test bed to conduct an inter-comparison starting in summer 2012 for such a performance evaluation. The goal of the inter-comparison is to compare target instruments to high accuracy radiation monitoring instruments (references) from the BSRN Payerne site. The BSRN reference instruments are directly traceable to the World Radiometric Reference [3] established by the World Radiation Center at Davos, Switzerland. This performance evaluation will allow verifying that they fulfill the requisites of the solar energy sector. This instrument performances assessment represents a first step towards establishing Standard Operating Procedure (SOP) for their use. These subsequent tasks (SOP, certification definition, etc.) should be carried out in collaboration with relevant official certification committee, instrument users in the solar energy sector and the IEA. The inter-comparison will include an intensive operation period (IOP) followed by a long-term performance evaluation period with a subset of the tested instruments. The instrument performance will be evaluated with respect to the reference for: a) direct normal irradiance, b) diffuse irradiance and c) global irradiance. The performance evaluation will evaluate statistical estimators for systematic (bias) and statistical errors (RMSE). The dependence of bias and RMSE with respect to solar zenith angle will be studied for evaluating the quality of the correspondence between the diurnal cycles as measured by the tested instruments and the references. Dependence on other pertinent explanatory parameters will also be evaluated.

2.1.2.2 WIND ENERGY

Meteorological wind measurements are essentially of three types. In-situ wind measurements with anemometers mounted on masts, radiosonde measurements and remote sensing measurements such as wind profilers and more cost-effective and portable wind lidars. Currently, in situ wind sensors installed on masts are the only fully recognized measurement reference when assessing the wind conditions at a given site with the purpose of installing a new wind turbine. Over the last years the wind lidar technology has shown its potential and is becoming mature, but “official standards” for it are lacking so far. Considering the state of development of such a remote sensing technique and the strong growth in the wind energy sector the development of standards such as ISO standard for wind lidars is an activity that would benefit the wind energy sector and where COST Action WIRE can provide a significant contribution. Meanwhile (Status 12/2011) a new ISO New Work Item has been initiated under the title ISO 28902 Part 2 "Air quality – Environmental meteorology – Part 2: Ground-based remote sensing by Doppler wind lidar" Similarly to solar energy, requirements from the wind energy sector for wind sensors are being defined. But in this case it is built on the substantial work already performed in collaborative projects such as SafeWind (www.safewind.eu), ANEMOS.plus (www.anemos-plus.eu), SEEWIND (www.seewind.org) and activities within COST WIRE. These requirements should be defined not only for wind anemometers on masts for which standards already exist, but also for remote sensing systems. For such systems, activities such as the ISO New Work Item 28902 Part 2 "Air quality – Environmental meteorology – Ground-based remote sensing by Doppler wind lidar" is particularly pertinent and COST WIRE should maintain collaboration with it.

2.1.2.2.1 PERFORMANCE EVALUATION FOR WIND ENERGY

5 See also Appendix A and B

The cup-anemometer is the traditional and generally recognized instrument for wind measurements in connection with wind energy assessment studies. Although there exists several other instruments for measurements of wind speed such as sonic-anemometers, radiosondes, wind profilers, acoustic sodars and wind lidars, mast mounted cup- anemometer is presently the only recognized one for use in wind energy assessment studies. The performance requirements for a cup-anemometer are described in the IEC61400-12 standard, in addition there are specific requirements set-out for wind speed calibration in the MEASNET guidelines. The calibration is performed in a wind tunnel with strict specifications for the performance of the wind tunnel and a round robin exercise (circulating a cup- anemometer between MEASNET approved institutions for comparison of calibration).

Wind lidars are a promising new development for measurements of wind profiles, but there is not yet a standard for their performance. Although comparisons of wind lidar and cup-anemometer measurements at 100 meters looks extremely promising , the wind lidar has not yet reached the level of confidence among the users that in general terms will make it applicable for wind energy assessment studies. A main shortcoming is presently the recommended requirement of an 85% recovery rate for the data, which seems to be a problem to fulfill. This shortcoming might well be rectified in the future.

Comparisons between wind –lidar and cup-anemometer mast measurements from masts have been performed at the Danish Test Station for Wind Turbines at Høvsøre, a coastal location in the western part of Denmark, but it has been limited to a height of 100 meters. A new test station at Østerild in the western coast of Denmark is under construction. It will be equipped with several 250 meters tall well instrumented meteorological masts. It is planned to perform comparison of wind lidars and mast measurements of wind speed and direction at this new site, and in this way contribute to the establishment of a calibration procedure for wind lidars.

2.1.2.3 POTENTIAL CONTRIBUTION OF THE COST ACTION ES1002 WIRE

The COST Action ES1002 WIRE has a pioneer role to play in the characterization of measurement techniques tuned to the energy sector. Since it involves partners active in weather forecasting and modeling, measurements of meteorological parameters and energy production, it is ideally suited for identifying parameters of interest, defining the observational requirements of the energy sector, and characterizing the performance of measuring instruments. Particularly, instrument performance evaluation such as the one that will be initiated in summer 2012 at the Payerne BSRN station for solar radiation measurement and similar activities in preparation for wind measurements (contributions to ISO 28902 Part 2 "Air quality – Environmental meteorology – Part 2: Ground-based remote sensing by Doppler wind lidar") are the starting point for defining the regulatory framework encompassing meteorological methods and products tuned for the energy sector.

Reference to §2.1.2:

[1] Joint Committee for Guides in Metrology, 2008, ‘Evaluation of measurement data — Guide to the expression of uncertainty in measurement’. Bureau International des Poids et Mesures, pp. 120 [2] International Standards for Business, Government and Society, List of ISO technical committees, TC 180/SC 1, http://www.iso.org/iso/standards_development/technical_committees/list_of_iso_technical_committees/iso_tec hnical_committee.htm?commid=54024 (accessed 20/02/2012) [3] Fröhlich, C., 1977, 'World Radiometric Reference', in: WMO/CIMO Final Report, WMO No. 490, 97-110.

3 RESEARCH AND DEVELOPMENT: THE EUROPEAN APPROACHES

This chapter deals with the achievements and difficulties of the forecasting of renewable energy production delivered by wind and solar energy systems. It will show that these in-situ applications (power plants or production sites) are only one component of the problems lying ahead, and that the integration in the existing or future electrical networks will be an even greater challenge. It demonstrates that a dual approach is necessary – production and integration - if the penetration of intermittent renewable energies such as wind and solar energies is to increase to levels comparable to other forms of power production less dependent on the weather conditions.

3.1 WIND ENERGY6

3.1.1 OVERVIEW

There are two main sources to learn about short-term forecasting of wind power, the report by Giebel et al. [1] produced first for the ANEMOS project, and then updated for the ANEMOS.plus and SafeWind projects, and the report from Argonne National Laboratory, written by Monteiro et al. [2]. The Giebel et al. report focuses on the scientific approaches and contains many conference papers, while the Monteiro et al. report has a detailed market overview of currently available commercial models and closes with the integration of wind power forecasts into the unit commitment process (the process of scheduling power plants, see 3.3.1) especially in the US. This chapter relies heavily on the Giebel et al. report. On the first half page, Giebel et al. point to 13 further overview papers on short- term prediction. The report also goes into detail on NWP (Numerical Weather Prediction) and Ensemble methods.

One of the largest challenges of wind power, as compared to conventionally generated electricity, is its dependence on the availability of the wind, and therefore its exposure to the wind variability. This behaviour happens on all time scales, but two of them are most relevant: one is for the turbine control itself (from milliseconds to seconds), and the other one is important for the integration of wind power in the electrical grid, and therefore determined by the time constants in the grid (from minutes to weeks). Turbine control is out of scope of this overview, as it involves mainly advection of a wind field measured a few seconds before it hits the turbine, usually using a lidar in the nose of the turbine, and therefore is qualitatively different from the rest of the approaches mentioned here.

One can distinguish the following types of applications:

• Optimisation of the scheduling of conventional power plants by functions such as economic dispatch etc. The prediction horizons can vary between 3-10 hours depending on the size of the system and the type of conventional units included (i.e. for systems including only fast conventional units, such as diesel generator sets or gas turbines, the horizon can be below 3 hours). Only a few fully integrated on-line applications of this type exist today. Typically, these systems are used for smaller or isolated power systems, like island systems, though the optimisation for larger systems as in Ireland was evaluated, e.g. in the ANEMOS.plus project [3]. • Optimisation of the value of the produced electricity in the market. Such predictions are required by different types of end-users like utilities, TSOs (Transmission System Operators), Energy Service Providers, IPPs (Independent Power Producers), energy traders etc., and for different functions such as unit commitment, economic dispatch, dynamic security assessment, participation in the electricity market, etc. Most forecasting in Europe is concerned with the time scale given by the electricity markets, which in most European countries ranges from 0 to 48 hours. • Allocation of reserves based on the expected wind power feed. This aims at system security and is done for instance in Ireland [4]. • Additionally, even longer time scales would be interesting for the maintenance planning of large power plant components, wind turbines or transmission lines. However, the accuracy of weather predictions decreases strongly when looking at 5-7 days in advance, and such systems are only starting to appear [4], [6], [7]. As Still [8] reported, shorter horizons can also be considered for maintenance, when it is important that the crew can safely return from the offshore turbines in the evening7. The north-western German Distribution System Operator (DSO) EWE8 [9] is integrating wind forecasts into transformer maintenance routines to assess the line loading of the remaining rerouted electricity flows.

6 Lead author: Gregor Giebel 7 The German Offshore Test Field Alpha Ventus had an incident like this in December 2009, when 11 workers were trapped for two days in a storm on the turbines in the North Sea. 8 Energieversorgung Weser-Ems AG

3.1.1.1 THE MODEL CHAIN In general, the short-term prediction models can either use NWP model input or not. Whether the inclusion of a NWP model is beneficial depends on the time horizon functional to the application. Typically, prediction models using NWP forecasts outperform statistical approaches based on time series analysis after approx. 3-6 hours look- ahead time. Therefore, all models employed by utilities use this approach.

When using NWP inputs for short-term power prediction, the physical and the statistical approaches may be applied. In most operational and commercial models, a combination of both is used, as indeed both approaches can be needed for successful forecasts. In short, the physical models try to use physical considerations as long as possible to reach to the best possible estimate of the local wind speed before using Model Output Statistics (MOS) or different relatively simple statistical techniques to reduce the remaining error. Statistical models in their pure form try to find the relationships between a wealth of explanatory variables including NWP results, and online measured power data, usually employing recursive techniques. Often, black-box models like advanced Recursive Least Squares or Artificial Neural Networks (ANN) are used. The more successful statistical models actually employ grey-box models, where some knowledge of the wind power properties is used to tune the models to the specific domain. Some of the statistical models can be expressed analytically, some (like ANNs) cannot. The statistical models can be used at any stage of the modelling, and often combine various steps into one.

If the model is formulated rather explicitly, as is typical for the physical approach, then the stages are the calculation of the wind at the site and height of the turbine, the so-called downscaling, then the conversion to power of that wind using some power curve, and the upscaling of the single turbine results to regional or national aggregates:

Downscaling

The wind speed and direction from the relevant NWP level is scaled to the hub height of the turbine. This involves a few steps, first finding the best-performing NWP level, often the wind speed at 10 m a.g.l. or at one of the lowest model or pressure levels. The 10 m wind speed and direction is a standard product every meteorological institute delivers. However, with the advent of the large multi-MW class of turbines, the hub height is close to 100m, which is why ECMWF (the European Centre for Medium Range Weather Forecast in Reading, UK) now also hands out a wind speed at 100 m as standard.

The NWP model results can be obtained for the geographical point of the wind farm or for a grid of surrounding points. In the first case the models could be characterised as “advanced power curve models”, in the second case as a “statistical downscaling” model. LocalPred for example uses principal component analysis and artificial intelligence techniques from the surrounding NWP grid points [10],[11].

The next step is the so-called downscaling procedure itself. The physical approach uses a meso- or microscale model for the downscaling, and both can be run for various cases in a look-up table approach. The differences between the two reside in the domain size, resolution, parameterizations and solvers. One of the reasons for using microscale models might be their ability to resolve scales down to metres depending on the terrain and on the computer power. .

However, since the microscale models are only capable of modelling local conditions, changing in large scale features cannot be forecasted therefore they either can be run for stationary conditions or coupled to large scale models if run for long time.

Conversion to power

The downscaling process yields a wind speed and direction for the turbine hub height. This wind is then converted to power with a power curve. The use of the manufacturers’ power curve is the easiest approach, although research

from a number of groups has shown it advantageous to statistically estimate the power curve from the forecasted wind speed and direction and measured power, leaving out the detailed steps described in the Upscaling section. To get the wind farm power, the power of the individual turbines has to be added and the wake effects between them subtracted. In this case, the estimation of the power curve should be for the total wind farm power versus the wind speed and direction of the NWP model.

Depending on forecast horizon and availability, measured power data can be used as additional input. In most cases, actual data is beneficial for improving the residual errors in a MOS approach. If online data is available, then a self- calibrating recursive model is highly advantageous for all horizons, but of course especially for the very short-term prediction. This is part of the statistical approach. However, sometimes only offline data is available, with which the model can be calibrated in hindsight. In recent years, a number of system operators (notably AEMO, the Australian Electricity Market Operator) have demanded to get online data from wind farms specifically to be used in their online prediction tools [15].

Upscaling

If only one wind farm is to be predicted, then the model chain stops here. Since utilities usually want a prediction for the total area they service, the upscaling from the single results to the area total is the last step. If all wind farms in an area were to be predicted, this would involve a simple summation. However, since practical reasons forbid the prediction for thousands of wind farms, some representative farms are chosen to serve as input data for an upscaling algorithm. Helpful in this respect is that the error of distributed farms is reduced compared to the error of a single farm (see also Figure 6).

Not all short-term prediction models involve all steps and/or all types of input. In the early days of forecasting (1970ies), NWP data was not so widely available, therefore the first approaches were done with time series analysis techniques. But in an age where at least GFS forecasts from the USA are just a download away, there is no real incentive to not use it. Leaving out a few steps can be an advantage in some cases. For example, Prediktor [16] is independent of online data, and can bring results for a new farm from day 1, while the advanced statistical models need older data to learn the proper parameterisations. However, this is bought with a reduced accuracy for the very short-term horizons. Landberg [17] has shown that a simple NWP + physical downscaling approach is effectively linear, thereby being very easily amenable to MOS improvements – even to the point of overriding the initial physical considerations.

The opposite is a direct transformation of the input variables to wind power. This is done by the use of grey- or black-box statistical models that are able to combine input such as NWPs of speed, direction, temperature etc. of various model levels together with on-line measurements such as wind power, speed, direction etc. With these models, even a direct estimation of regional wind power from the input parameters in a single step is possible. Whether it is better for a statistical model to leave out the wind speed step depends on a number of aspects like the availability of data or the representativity of the wind speed and power for the area of the wind farm or region being forecasted.

The optimal model is a combination of both, using physical considerations as far as necessary to capture the air flow in the region surrounding the turbines, and using advanced statistical modelling to make use of every bit of information given by the physical models.

3.1.1.2 TYPICAL FORECAST ERRORS

Most of the errors on wind power forecasting stem from the NWP model. There are two types of error: level errors and phase errors. Consider a storm front passing through: a level error misjudges the severity of the storm, while a phase error misplaces the onset and peak of the storm in time. While the level error is easy to get hold of using standard time series error measures, the phase error is harder to quantify, although it has a determining impact on the

traditional error scores. A conundrum for forecasters is that higher resolution forecasts tend to capture more of the variability, but if there is just a slight phase error, the traditional error scores as explained in the following will be worse than with a very smooth forecast, even though the operator is probably more fond of the more “realistic” looking forecast.

Figure 2: Typical errors introduced by the NWP. Source: Möhrlen [18].

The ISET (Institut für Solare Energieversorgungstechnik e.V., Kassel, Germany, now Fraunhofer IWES) was the first short-term forecasting provider for transmission system operators in Germany. In a widely cited paper for the EWEC 2006, B. Lange et al. [19] presented the following plot for the accuracy of the next-day forecast in the E.On control zone. They state the main reasons for the improvement were (i) taking into account the influence of atmospheric stability into the models which led to a reduction in forecast error (RMSE) by more than 20% for the example of one German TSO control zone, and (ii) a combination of different models, both for forecasting methods as well as for NWP models. The comparison of the mean RMSE of a wind power forecast for Germany obtained with the WPMS (Wind Power Management System) based on ANN with input data from three different NWP models and with a combination of these models showed a decrease in RMSE from approx. 6% to 4.7%.

Note that their competitor, energy&meteo systems, claims a forecasting RMSE of below 5% for the day-ahead forecast for all of Germany in 2008 [20], which also the IWES has achieved [21].

A similar plot, though constrained to the last two years, was shown by Krauss et al. [22] for the EnBW TSO area. They show the monthly accuracy of three different forecasting systems for the aggregate error, and conclude that there are significant changes in forecast accuracy from month to month, and that the ranking of the three models changes from month to month as well.

Figure 3: The development of the forecast error during the last years in the E.On Netz area. The numbers in square brackets are references from Lange et al. [19].

3.1.2 VERY SHORT-TERM PREDICTION MODELS

For very short horizons, the relevant time scales are given by:

. the mechanics of the wind turbine: typically the generator, gearbox, yaw mechanism and most of all the (blade) pitch regulation. The time scales involved are in the order of turbulence, i.e. seconds. The purpose is the active control of the wind turbines. Wind on those time scales is inherently non-stationary (compare also the excursion on why wind is non-stationary in [23]), and can best be forecasted with a Lidar staring into the wind and a simple advection scheme of the measured wind field a few seconds ahead the rotor. . the type of the power system into which the wind turbines are integrated. As mentioned in the introduction in small or medium isolated systems the relevant time scale is given by the type of conventional units (“fast” or “slow”) and the functions for which the forecasts are required (i.e. for economic dispatch horizons can be 10 minutes to 1 hour while for unit commitment they can be a few hours head). It is typical for smaller island systems to consist of Diesel generators with quite short time scales. . the type of market the power system is operating by. In some countries (e.g. Australia and parts of the US), the main market is a 5-min ahead market, which means that also the wind power predictions for 5 minutes ahead are of major importance.

The typical approach is to use time series analysis techniques or neural networks. The easiest technique (and the typical benchmark) is persistence, essentially stating that the future wind is just like the wind now.

While there had been attempts to forecast wind speeds before, the first paper considering wind power forecasts came from Brown, Katz and Murphy in 1984 [24]. In retrospect, it is surprising how complete the paper already was, using a transformation to a Gaussian distribution of the wind speeds, forecasting with a AR (AutoRegressive) process, upscaling with the power law (but discussing the potential benefit of using the log law), and then predicting power using a measured power curve. Additionally, the removal of seasonal and diurnal swings in the AR components is discussed, alongside prediction intervals and probability forecasts.

Artificial Neural Networks (ANN) are another family of models that use data from online measurements as input. Most groups in the field have used them, but despite their scientific merits in improvements over plain persistence, they did not catch on. The improvements attainable were usually deemed not enough to warrant the extra effort in training the neural networks. This assessment ends different if both measured power and NWP input are taken into account.

Many other authors tried ARMA, ARIMA and related models plus many varieties of neural networks to beat persistence, and often did not get more than 10% improvement. See the whole chapter on time series models in Giebel et al. [1] for more information.

To be able to improve further, Pinson et al. [25] found that wind power and especially wind power variability from large offshore wind farms (Horns Rev and Nysted) occur in certain regimes, and therefore tested “regime-switching approaches relying on observable (i.e. based on recent wind power production) or non-observable (i.e. a hidden Markov chain) regime sequences” for a one-step forecast of 1-min, 5-min and 10-min power data. “It is shown that the regime-switching approach based on MSAR models significantly outperforms those based on observable regime sequences. The reduction in one-step ahead RMSE ranges from 19% to 32% depending on the wind farm and time resolution considered.”

3.1.3 NUMERICAL WEATHER PREDICTION

The main error in the final forecast comes from the meteorological input. For example, Sanchez et al. [20] show that the Spanish statistical tool Sipreolico run with on-site wind speed input has a much higher degree of explanation than HIRLAM forecasts. This means that given a representative wind speed, Sipreolico can predict the power quite well. It is the wind speed input from the NWP model that is decreasing the accuracy significantly. Therefore, it is logical to try to improve the NWP input in order to come up with significant improvement in forecasting accuracy.

Figure 4: The error comes from the NWP. The figure shows the difference in degree of explanation between Sipreolico run with HIRLAM input (from an older version of the Spanish HIRLAM) and Sipreolico run with on-site wind speed input. Source: Sanchez et al. [26].

Various global forecasting NWPs exist, designed to predict large scale synoptic weather patterns. But the increase in computer resources during the next years will allow the global models to overtake the current role of the limited area models (LAM) down to about 10 km horizontal resolution. At the moment most countries run their models for overlapping European areas at 12-7 km grid resolution. In the next few years they will move towards 4-1 km grid resolution and therefore will not run an intermediate nested European grid area anymore. They plan to directly nest their very high resolution models, which then will cover only the national area, into a global model at 25km or less. For very high resolution requirements of a European wide short range NWP coverage a need arises for close cooperation and exchange of NWP products. There are already operational suits running at very high resolution in most of the European weather services. Figure 4 shows some examples of model domains:

Figure 5: Very high resolution model domains from left to right: AROME (Meteo France, 41 vertical layers), COSMO-DE (DWD, 50 vertical layers), UM-4km (grey shaded area, UK Met Office, 70 vertical layers).

A large effort to the aim of meteorological forecasts for wind energy purposes has also been made by the original ANEMOS project. A report [27] details some work related to downscaling techniques with microscale, mesoscale and CFD models. If possible from a computational point of view, two-way nesting between domains is clearly preferred. While one group using mostly physical modeling reported increased accuracy down to two kilometers grid spacing, another one using an advanced statistical model claimed no improvement when going from 9 to 3 km grid spacing. This is probably due to the fact that the forecasted time series become more “realistic” when increasing horizontal resolution, in the sense that the ups and downs of the time series have similar amplitude to the original series in the high frequency domain. However, this means a higher potential for phase errors, so for the usual RMS error or MAE the error goes up. Increasing the horizontal resolution beyond the resolution of the terrain database is fairly useless. On the other hand, increasing the vertical resolution in the lowest, say, 200m of the atmosphere improved the results in all cases. The report closed with the following recommendations: “If you have a site in complex terrain, where you even after using an advanced MOS are not happy with the forecasts, then try to use higher resolution modeling. In many cases and with a large number of approaches, the models can improve the NWP results. When setting up a model yourself, make sure to use the best terrain DB available (e.g. SRTM data), and try to get good NWP input data. Set up the model to have good vertical resolution, and reasonable horizontal resolution. Find out for yourself what “reasonable” means in this context. Use a MOS. Use insights gleaned from high- resolution modeling to decide which parameters to employ in the MOS. In any case, setting up a model from scratch will take a long time before one is familiar with the model and its quirks, so do not plan on having a solution up and running immediately.”

Since the increase in horizontal grid resolution did not necessarily yield improved forecast error scores, it can be argued that it is better to use the computer resources for ensemble forecasts, where the initial state of the forecast is disturbed so as to get diverging weather developments with the same probability of occurrence. In Europe seven operational limited-area ensembles are running at the large meteorological centers (and the one by WEPROG). Some of these systems will be stored in a central database within the TIGGE-LAM project, the Limited Area Model component of TIGGE (see http://www.smr.arpa.emr.it/tiggelam/ ).

3.1.4 SHORT-TERM PREDICTION MODELS

This section deals with models using NWP data as input, which includes all the commercial models. Probably the earliest model was developed by McCarthy [28] for the Central California Wind Resource Area. It was run in the summers of 1985-87 on a HP 41CX programmable calculator, using meteorological observations and local upper air observations. The program was built around a climatological study of the site and had a forecast horizon of 24 hours. It forecast daily average wind speeds with better skill than either persistence or climatology alone.

In 1990, Landberg and Troen [29][30] developed a short-term prediction model, now known as Prediktor, based on physical reasoning similar to the methodology developed for the European Wind Atlas [31]. The idea is to use the wind speed and direction from a NWP, transform this wind to the local site, use the power curve and finally modify this with the park efficiency. Note that the statistical improvement module MOS can either be applied before the transformation to the local wind, before the transformation to power, or at the end of the model chain to operate on the power itself. A combination of all these is also possible.

The Wind Power Prediction Tool (WPPT) has been developed by the Institute for Informatics and Mathematical Modelling (IMM) of the Technical University of Denmark. In 2006, the original developer Torben Skov Nielsen together with Henrik Madsen and Henrik Aalborg Nielsen founded the DTU spin-off company ENFOR, which now stands for all commercial activity with the model. WPPT has been running operationally in the western part of Denmark since 1994 and in the eastern part since 1999. The model uses adaptive recursive least squares estimation of the parameters of conditional parametric models to find the best connection between the NWP predicted wind speeds for the site and the measured power for each forecast horizon. A central part of this system is the statistical models for short-term prediction of the wind power production in wind farms or areas. For on-line applications it is advantageous to allow the function estimates to be modified as data becomes available. Furthermore, because the system may change slowly over time, observations should be down-weighted as they become older. For this reason a time-adaptive and recursive estimation method is applied. Depending on the available data the WPPT modeling system employs a highly flexible modeling hierarchy for calculating predictions of the available wind power from wind turbines in a region. For a larger region this is typically done by separating the region into a number of sub- areas. Wind power predictions are then calculated for each sub-area and hereafter summarized to get a prediction for the total region. The predictions for the total region are calculated for a number of reference wind farms using online measurements of power production as well as numerical weather predictions as input.

A rather similar approach to Prediktor was developed at the University of Oldenburg [32], called Previento. A good overview over the parameters and models influencing the result of a physical short-term forecasting system has been given by Mönnich [33]. He found that the most important of the various sub-models being used is the model for the atmospheric stability. The use of MOS was deemed very useful. However, since the NWP model changed frequently, the use of a recursive technique was recommended. A large influence was found regarding the power curve. The theoretical power curve given by the manufacturer and the power curve found from data could be rather different. Actually, even the power curve estimated from data from different years could show strong differences. The latter might be due to a complete overhaul of the turbine. The largest influence on the error was deemed to come from the NWP model itself. In 2004, the two principal researchers behind Previento, Matthias Lange and Ulrich Focken, left the University to form energy & meteo systems, a company which had good success from the start and has now over 20 employees. Their work on the weather dependent combination of models is also published in [20] or [34]. In essence, principal component analysis identifies between 5 and 8 different weather types, and the model parameters are optimized according to weather type.

ARMINES and RAL have developed work on short-term wind power forecasting since 1993. Initially, short-term models for the next 6-10 hours were developed based on time series analysis to predict the output of wind farms. In the frame of the project CARE [35], more advanced short-term models were developed for the wind farms installed in Crete. In the follow-up project MORE-CARE, ARMINES developed models for the power output of a wind park for the next 48/72 hours based on both on-line SCADA and Numerical Weather Predictions (meteorological forecasts). The wind forecasting system of ARMINES integrates:

• short-term models based on the statistical time-series approach able to predict efficiently wind power for horizons up to 10 hours ahead. • longer-term models based on fuzzy neural networks able to predict the output of a wind farm up to 72 hours ahead. These models receive as input on-line SCADA data and numerical weather predictions [36].

• combined forecasts: such forecasts are produced from intelligent weighting of short-term and long term forecasts for an optimal performance over the whole forecast horizon.

The ISET (Institut für Solare Energieversorgungstechnik, now the main part of the Fraunhofer Institut für Windenergie und Energiesystemtechnik IWES) has since 2000 operatively worked with short-term forecasting, using the DWD model and neural networks. It came out of the German federal monitoring program WMEP (Wissenschaftliches Mess- und EvaluierungsProgramm) [37], where the growth of wind energy in Germany was to be monitored in detail. Their first customer was E.On, who initially lacked an overview of the current wind power production and therefore wanted a good tool for nowcasting [38]. The Artificial Neural Network (ANN) employed by them also provides for an area power curve. The WPMS runs at E.On since 2001, at RWE since June 2003, for Vattenfall Europe since the end of 2003, and in a variety of other places as well [39]. A version for two hours horizon has been developed for National Windpower in the UK. For the E.On total area, they claim RMSE values of 2,5% for 1h horizon (persistence would be 3,3%), 5,2% (persistence: 7,3% ) at 3h, 6% (persistence: 9%) at 4h, and reach the error of a purely NWP based prognosis (7,5%) at 7h horizon.

The Sustainable Energy Research Group (SERG) in University College Cork (UCC) has been researching and developing wind power forecasting methodologies based on ensemble forecasts in the years 2002-2006, see e.g. [40][41][42][43][44]. An operational forecasting system was developed by the principle researchers in UCC and brought to life in 2004 by WEPROG (Weather and wind Energy PROGnosis), which was founded in 2003. WEPROG's MSEPS (Multi-Scheme Ensemble Prediction System) contains a 2-step power prediction module. In the first step a physical reference power is computed and in a second step, the reference power is localised statistically and with the help of weather classes defined by the ensemble weather input. eWind is an US-American model by TrueWind, Inc (now AWS TruePower) [45]. The current iteration of eWind uses ARPS, MASS and WRF, fed by the global models GFS, GEM and ECMWF, to yield an ensemble of 9 different model runs [46]. For the average prediction of 6 wind farms in Europe, their “results reveal that the ensemble prediction outperforms the accuracy of [...] the MOS method applied to single NWP models, achieving between a 20 and 30 % of improvement during the first three days of prediction.” Zack [47] presented their high resolution atmospheric model to operate in a rapid update cycle mode, called WEFRUC – Wind Energy Forecast Rapid Update Cycle. The model assimilates different types of data available in the local-area environment of a wind plant such as remotely sensed data, which is the starting point for a short-term simulation of the atmosphere. So, the atmospheric simulation produced by the physics based model is incrementally corrected through the use of the measured data as it evolves. Their update cycle is 2 hours.

The strong wind energy growth in Spain led Red Eléctrica de España (the Spanish TSO) to have the Sipreólico tool developed by the University Carlos III of Madrid [48]. The tool is based on Spanish HIRLAM forecasts, taking into account hourly SCADA data from 80% of all Spanish wind turbines [49]. These inputs are then used in adaptive non-parametric statistical models, together with different power curve models. These 9 models are recursively estimated with both a Recursive Least Squares (RLS) algorithm and a Kalman Filter. The results of these 18 models are then used in a forecast combination [50], where the error term is based on exponentially weighted mean squared prediction error with a forgetting factor corresponding to a 24-h memory.

LocalPred and RegioPred [10] are a family of tools developed by Martí Perez et al. (CENER). Originally, it involved adaptive optimization of the NWP input based on principal component analysis, time series modeling, mesoscale modeling with MM5, and power curve modeling. They could show for a case of rather complex terrain near Zaragoza, Spain, that the resolution of HIRLAM was not good enough to resolve the local wind patterns [51]. The two HIRLAM models in Spain were at the time running on a 0.5°x0.5° and 0.2°x0.2° resolution. Successive research and development carried out at CENER [52] has transformed LocalPred into a multi model wind power forecasting system. In its current form, an ensemble forecasting model takes MM5, Skiron and the ECMWF model as NWP inputs for learning machine techniques such as cluster or support vector machines [11]. The final prediction is offered by an adaptive model that combines all the individual inputs.

GL Garrad Hassan [53] has a forecasting model called GH Forecaster, based on NWP forecasts from the UK MetOffice. It uses "multi-input linear regression techniques" to convert from NWP to local wind speeds. For T+24h, they reach 35-60% improvement over persistence.

3Tier Environmental Forecast Group [54] works with a nested NWP and statistical techniques for the very short term in the Pacific Northwestern US. They show performance figures in line with most other groups in the field.

In the Nordic countries, but also in Canada, icing of wind turbines can decrease the production as the turbines need to shut down, or as the aerodynamic efficiency is strongly reduced due to ice aggregation. The Winterwind conferences are specialized in icing predictions. Thompson [55] talked on the potential of WRF and current developments for direct icing forecasts. Landberg [56] showed an example of power curve degradation due to icing. Durstewitz [57] reported on the difficulties encountered in Germany. Heimo [58] presented the European COST Action 727 “Measuring and forecasting atmospheric icing on structures”.

3.1.5 UPSCALING

Many larger clients are more interested in the result for a region than for a single wind farm, e.g. an electrically defined region as for Transmission System Operators (TSOs) or a market region as for traders. In only very few cases, typically where wind power only took off in the last few years, there is online data available for all turbines in a region. In many cases though, like in Denmark, the production data for most wind turbines is only available from the accounting system for payments for the wind turbine owners, with a delay of up to a month. This means that for the purposes of an online forecast, it is useless (it can be used for a MOS system). Therefore, a correlation has to be found between a few wind farms delivering online data within a region, and the much later determined total regional production.

Since not all wind farms in a region see the same wind speed at the same time, and since the error made by the NWP is temporally and spatially distributed, the error for forecasting a region is smaller than the error for a single wind farm. In this context it is interesting to investigate the spatial correlations between both the wind power generation and the wind power forecasting errors, as it is the uncorrelated part of the error which generates the error improvement due to spatial smoothing.

The variability of an averaged time series, eg expressed as the relative standard deviation of this time series, depends on the respective variability of the single time series, and on the correlation between the various series. For wind power forecasting, there are two effects which reduce the forecast error for a region in comparison to the one of a single wind farm: the generation as such is already smoother for a region due to the uncorrelated frequencies of the single wind farm generation profiles, making it thereby more easily predictable, and the forecast errors are uncorrelated on an even smaller length scale. For the former issue, refer to the literature overview given by Giebel [59]. In most studies, the generation correlation decays on a length scale of about 750 km.

The methodically most relevant study on the subject was made by Lange [34] and Focken [60]. They applied power measurements on 30 wind farms in Germany to study the accuracy of the aggregated power output of wind farms distributed over given regions.

Figure 6: Spatial cross-correlation of prediction deviations for various prediction times based on German data for the years 1996–1999. For comparison the cross-correlation coefficients of the prediction (36 h) are also shown. All cross-correlation coefficients have been averaged over 25km bins. The figure is provided by M. Lange, energy & meteo systems GmbH.

One of the results of Lange and Focken’s studies is the calculated cross-correlations shown in Figure 5, using a prediction method based on NWP results. The German results exhibit significantly longer distances than the Danish results in [61]. Comparing the Danish and German results, they agree quite well for distances less than 100 km.

According to Focken et al. [62], the increased cross-correlation for increased forecast horizons might be due to the growing systematic errors for increasing forecast horizon which give rise to higher spatial correlations. For comparison the cross-correlation coefficients for the 36 h power prediction have been calculated in the same way and are shown in [62] as well.

Lange and Focken have also analyzed normalized standard deviations of forecast errors. The standard deviations are normalized with the rated power of the corresponding wind power. If an ensemble consists of a number N of wind farms, then the relative standard deviation σensemble of the ensemble forecast error can be calculated according to

1 N N σ = σ σ ensemble 2 ∑∑ rxy x y (2) N x=1 y=1

where σx is the relative standard deviation of the forecast error of wind farm x power, and rxy is the cross correlation coefficient between forecast errors on wind farms x and y.

Figure 7: Forecast error standard deviation ratio versus region size quantified by the region diameter. The horizontal line gives the expected error reduction for an area the size of Germany. The figure is provided by M. Lange, energy & meteo systems GmbH.

3.1.6 RAMP AND VARIABILITY FORECASTING

In the early days of wind power, installations in e.g. Denmark and Germany were small and well distributed. This led to a quite smooth wind power feed. In recent years though, especially in the new markets such as Australia, USA or Canada, wind farms (generally offshore) are installed in 100-150 MW or even larger blocks. This leads to a much larger possibility for quick variations, or ramps. Those make life difficult for the personnel in the control room, as the wind feed can suddenly decrease several GW, going far out of the bounds of the usual spinning reserve requirements.

Bonneville Power Administration [63] held a competition dedicated for ramp forecasting. The first results [64] indicated that for ramps, hourly predictions are not good enough, and shorter timings of the forecast lead to smaller deviations. However, as Focken [65] pointed out, in the subsequent Request for Proposals for a short-term prediction system, ramps are not mentioned at all. Focken (having been part of the ramp forecasting competition with his company energy & meteo systems) attributes this to the fact that a ramp does not have an action in the control room associated with it – “the operators don’t know what to do with a ramp forecast”. In the remainder of his talk he points out that the ramp forecast needs to be something separate from the usual RMS-optimized forecast, since this tends to be too smooth.

Xcel Energy currently has a project on ramp prediction together with NCAR and Vaisala. The Finnish measurement company thereby tries to get into the solutions market with the commercial offering of their RampCast product [66], based on a set of masts around an existing wind farm and aiming at 0-3 hours prediction horizon. From 3 to 60 hours or more, NCAR’s DICast [67] uses a Dynamic MOS (DMOS) to find the best inputs for the removal of bias between the nacelle wind speeds of every individual turbine and one of 30 different WRF and MM5 runs. The DMOS parameters are recalculated every week and are differentiated by model run time and lead time. Then, the individual forecasts are combined into a consensus forecast “analogous to the job done by a human who, once having removed

biases from individual models’ forecasts must combine them into a single final forecast.” The DMOS step outperforms the best predictor by about 5-10% of RMSE error, while the consensus step reduces the error further 10- 15%. For the ramp forecast on the 0-3 hour horizon, Haupt et al. [68] use a Variational Doppler Radar Analysis System (VDRAS) for the nowcasting of the wind field. For the example of one ramp in Colorado, they show the advantage of using regional measured data for the very short term forecast.

While ramp forecasting and variability forecasting bear some resemblance, the two are actually quite different. Variability forecasting refers to large amplitude, periodic changes in wind speed, and it is only recently that it has come into the sight of researchers. Davy et al. [69] defined an index of variability based on the standard deviation of a band-limited signal in a moving window, and developed methods to statistically downscale reanalysis data to predict their index. Amongst the important predictors of variability, they found planetary boundary layer height, vertical velocity and U wind speed component during the months June-September (southern hemisphere winter), and U-wind speed, geopotential height and cloud water for the months December-February (southern hemisphere summer).

Vincent et al. [70],[71] defined a variability index as the sum of all amplitudes occurring within a given frequency range based on an adaptive spectrum. They studied the climatological patterns in variability on time scales of minutes to 10 hours at the Horns Rev wind farm, and showed that there were certain meteorological conditions (mostly open cellular convection) in which the variability tended to be enhanced. For example, variability had a higher average amplitude in flow from sea than in flow from the land, often occurred in the presence of precipitation and was most pronounced during the autumn and winter seasons.

3.1.7 UNCERTAINTY

Spot predictions of the wind production for the next 48 hours at a single wind farm or at a regional/national level are a primary requirement for end-users. However, for an optimal management of the wind power production it is necessary to also provide end-users with appropriate tools for on-line assessment of the associated prediction risk. Confidence intervals are a response to that need since they provide an estimation of the error linked to power predictions. Essentially, two main methodologies for uncertainty forecasting have established themselves in the industry: statistical approaches working on single NWP forecasts, and uncertainties derived from ensembles of predictions.

While statistical models already have an estimate of the uncertainty explicitly integrated in the method, physical models need some additional processing to yield an uncertainty result as well.

Typical confidence interval methods, developed for models like neural networks, are based on the assumption that the prediction errors follow a Gaussian distribution. This however is often not the case for wind power prediction where error distributions may exhibit some skewness, while the confidence intervals are not symmetric around the spot prediction due to the form of the wind farm power curve. On the other hand, the level of predicted wind speed introduces some nonlinearity to the estimation of the intervals; e.g. at the cut-out speed, the lower power interval may suddenly switch to zero.

Bremnes [72] developed a probabilistic forecasting technique, estimating the different quantiles of the distribution directly. In [73], he describes his method of local quantile regression (LQR) in more detail, and shows that for a test case in Norway, Hirlam forecasts have a lower inter-quantile range than climatology, which means that the Hirlam forecasts actually exhibit skill. LQR Hirlam features about 10% better in economic terms than pure Hirlam forecasts, increasing the revenue from approx.75-79% of the ideal income (without any forecast errors) to approx.79-86%, depending on the horizon. However, his pure Hirlam forecasting did not have an upscaling or MOS step, so this might have worked in favor of LQR in the comparison. He proposed to use the method to reduce the large amount of information found in meteorological ensembles. The motivation for this was that he could show that the

economically optimal quantile was not the central (“best”) quantile, but one given by the relative prices of up- and down-regulation.

Pinson and Kariniotakis [74][75] propose a methodology for the estimation of confidence intervals based on the resampling approach. This method is applicable to both physical and statistical wind power forecasting models. The authors also present an approach for assessing on-line the uncertainty of the predictions by appropriate prediction risk indices (“Meteo-Risk Index”) based on the weather stability. They use a measure of the distance (or the similarity) of subsequent predictions in a poor-mans ensemble. The approach was verified using HIRLAM forecasts and data from 5 wind farms in Ireland.

In Denmark, DTU.IMM and Risø National Laboratory (now DTU Wind Energy) had a nationally funded three-year project [7] on the use of different kinds of ensembles for utility grade forecasting. Amongst others, the NCEP/NCAR and ECMWF ensembles were used, multi-model ensembles (with input from both DMI and DWD) were compared, and some methods for a good visual presentation of the uncertainty were researched. One main result [76] was the development of a technique to transform the quantiles of the meteorological distribution to the quantiles of the power forecast distribution. The resulting quantiles were sharp and skillful. The use of pure meteorological ensemble quantiles was shown to be insufficient, since the ensemble spread is not probabilistically correct. Even using the transformation it was not possible to get satisfactory outer quantiles (eg below 15% and above 90%), since the meteorological ensemble spread is not large enough. This is especially relevant for the first days of the ensemble runs. However, in practice this might be less of a problem, since the ensemble runs also needed around 17 hours to complete, therefore making the first day impossible to use operatively. The model was used in a demo application run for two Danish test cases, the Nysted offshore wind farm and all of the former Eltra area (the Denmark West TSO). The results were quite satisfactory, have a horizon of one week, and were used for maintenance scheduling of conventional power plants, for the weekly coal purchase planning and for trading on the Leipzig electricity exchange, which is closed over the weekends.

Taylor, McSharry and Buizza [77] create a calibrated wind power density from the ECMWF EPS system. “The resultant point forecasts were comfortably superior to those generated by the time series models and those based on traditional high resolution wind speed point forecasts from an atmospheric model.”

Pinson and Madsen [78] “describe, apply and discuss a complete ensemble-based probabilistic forecasting methodology” for the example case of Horns Rev as part of the Danish PSO research project “HREnsembleHR – High Resolution Ensemble for Horns Rev, funded by the Danish PSO Fund from 2006-2009 (see www.hrensemble.net). The forecasts from WEPROG's 75 member MSEPS ensemble are converted to power using the novel orthogonal fitting method. The single forecasts are then subjected to adaptive kernel dressing with Gaussian kernels, since “in theory, any probabilistic density may be approximated by a sum of Gaussian kernels”, meaning that the resulting probabilistic distribution can be “a non-symmetric distribution (and possibly multimodal), thus being consistent with the known characteristics of wind power forecast uncertainty”.

3.1.8 THE VALUE OF FORECASTING

Even though the necessity and advantages of wind power forecasting are generally accepted, there are not many analyses that have looked in detail into the benefits of forecasting for a utility. Partly this lack of analyses stems from the fact that a lot of data input and a proper time step model are needed to be able to draw valid conclusions. In recent years, a number of wind integration studies have undertaken the effort with data backing from typically the TSO. Many researchers though found a significant value in the forecast, comfortably surpassing the likely installation cost of the software.

The importance and impact of good forecasts was for example stated by Operations Manager Carl Hilger from Eltra (the antecessor of Energinet.dk [79]: “If only we improved the quality of wind forecasts with one percentage point, we would have a profit of two million Danish crowns.” Similar orders of magnitude are quoted infrequently by other

utilities or traders, but usually not for publication. For the Xcel Energy forecasting project, arguably the largest and most ambitious privately funded forecasting project to date, Parks [80] reported savings of 6 million US$ for one year alone for three different regions. This significantly exceeds their investment.

References to Wind Power Forecasting:

[1] Giebel G., Brownsword R., Kariniotakis G., Denhard M., Draxl C. The State-Of-The-Art in Short-Term Prediction of Wind Power A Literature Overview, 2nd Edition. Project report for the Anemos.plus and SafeWind projects. 110 pp. Risø, Roskilde, Denmark, 2011 [2] Monteiro, C., R. Bessa, V. Miranda, A. Botterud, J. Wang, and G. Conzelmann: Wind Power Forecasting: State-of-the-Art 2009. Argonne National Laboratory ANL/DIS-10-1, November 2009 (see also http://www.dis.anl.gov/projects/windpowerforecasting.html) [3] Meibom, P.: Stochastic scheduling: the experience of Ireland. Talk on the final workshop of the ANEMOS.plus project, Paris (FR), 29 June 2011 [4] Barry, D.: The Irish Experience. Talk on the Workshop for Best Practice in the Use of Short-term Forecasting of Wind Power, Delft (NL), 25 October 2006 [5] Moreno, P., L. Benito, R. Borén and M. Cabré: Short-Term Wind Forecast. Results of First Year Planning Maintenance at a Wind Farm. Poster presented on the European Wind Energy Conference and Exhibition, Madrid (ES), 16-20 June, 2003 [6] Giebel, G., L. Landberg, C. Bjerge, M.H. Donovan, A. Juhl, K. Gram-Hansen, H.-P. Waldl, T. Pahlke, J. Giebhardt, M. Rebbeck, R. Ruffle, O. Brady: CleverFarm - First results from an intelligent wind farm. Paper presented at the European Wind Energy Conference and Exhibition, Madrid, Spain, 16-19 June 2003 [7] Gregor Giebel (ed.), Jake Badger, Lars Landberg, Henrik Aalborg Nielsen, Torben Skov Nielsen, Henrik Madsen, Kai Sattler, Henrik Feddersen, Henrik Vedel, John Tøfting, Lars Kruse, Lars Voulund: Wind Power Prediction using Ensembles. Risø-R-1527, September 2005 [8] Still, D., B. Grainger: Demanding Seas – The UK’s First Offshore Wind Farm. Proceedings of the European Wind Energy Conference, Copenhagen, Denmark, 2-6 June 2001, pp. 169-171, ISBN 3-936338-09-4. Note: This statement was done during the talk. [9] Focken, U., J. Jahn, and M. Schaller: Transformer Congestion Forecast Based on Highly Localized Wind Power. Talk on the 8th International Workshop on Large-Scale Integration of Wind Power into Power Systems as well as on Transmission Networks for Offshore Wind Farms, Bremen (DE), 14-15 October 2009 [10] Martí Perez, I.: Wind Forecasting Activities. Proceedings of the First IEA Joint Action Symposium on Wind Forecasting Techniques, Norrköping, Sweden, December 2002, pp. 11-20. Published by FOI - Swedish Defence Research Agency. [11] Frías, L., E.Pascal, U.Irigoyen, E.Cantero, Y.Loureiro, S.Lozano, PM. Fernandes, I.Martí: Support Vector Machines in the wind energy framework. A new model for wind energy forecasting. Proc. Of the 2009 European Wind Energy Conference EWEC'09 Marseille (FR), 2009 [12] Pielke, R.A., and E. Kennedy: Mesoscale terrain features. January 1980 report Uva Env. Sci meso 1980-1 University of Virginia Dept of Environmental Science. [13] Young, G. S., and R.A. Pielke: Application of terrain height variance spectra to meso scale modelling. J. Atmos. Sci. 40, pp 255-2560, 1983 [14] Skamarock, W.C.: Evaluating Mesoscale NWP Models Using Kinetic Energy Spectra. Monthly Weather Review 132, p. 3019-3032, December 2004 [15] Juneja, D.: Australian Wind Energy Forecasting System Overview. Talk on the 3rd Workshop on Best Practice in the Use of Short-term Prediction of Wind Power, Bremen (DE), 13 October 2009 [16] Landberg, L.: Short-term Prediction of the Power Production from Wind Farms. J. Wind Eng. Ind. Aerodyn. 80, pp. 207-220 (1999) [17] Landberg, L.: A Mathematical Look at a Physical Power Prediction Model. Wind Energy 1, pp. 23-28 (1998).

[18] Möhrlen, C.: First experiences of the new EEG trading rules in Germany and introduction of an ensemble based short-term forecasting methodology for intra-day trading. Talk on the 3rd Workshop on Best Practice in the Use of Short-term Prediction of Wind Power, Bremen (DE), 13. 10. 2009 [19] Lange, B., K. Rohrig, B. Ernst, F. Schlögl, Ü. Cali, R. Jursa, and J. Moradi: Wind power prediction in Germany – Recent advances and future challenges. European Wind Energy Conference and Exhibition, Athens (GR), 27.2.-2.3. 2006 [20] Lange, M., U. Focken, R. Meyer, M. Denhardt, B. Ernst, F. Berster: Optimal Combination of Different Numerical Weather Models for Improved Wind Power Predictions. 6th International Workshop on Large-Scale Integration of Wind Power and Transmission Networks for Offshore Wind Farms, Delft, 2006 [21] Wessel, A., R. Mackensen, B. Lange: Development of a shortest-term wind power forecast for Germany including online wind data and implementation at three German TSO. Talk on the 3rd Workshop for Best Practice in the Use of Short-term Forecasting, Bremen (DE), 13 Oct 2009 [22] Krauss, C, B. Graeber, M. Lange, U. Focken: Integration of 18GW Wind Energy into the Energy Market – Practical Experiences in Germany. Workshop on the Best Practice in Short-term Forecasting, Delft (NL), 25 October 2006 [23] Vincent, C., G. Giebel, P. Pinson, and H. Madsen: Resolving Non-Stationary Spectral Information in Wind Speed Time Series using the Hilbert-Huang Transform. Journal of Applied Meteorology and Climatology 49(2), pp. 253-267, February 2010 [24] Brown, B.G., R.W. Katz, and A.H. Murphy: Time Series Models to Simulate and Forecast Wind Speed and Wind Power. Journal of Climate and Applied Meteorology 23(8), pp. 1184-1195, August 1984 [25] Pinson, P., L.E.A. Christensen, H. Madsen, P.E. Sørensen, M.H. Donovan, and L.E. Jensen: Regime-switching modelling of the fluctuations of offshore wind generation. Journal of Wind Engineering and Industrial Aerodynamics 96(12), pp. 2327– 2347, 2008 [26] Sánchez, I., J. Usaola, O. Ravelo, C. Velasco, J. Domínguez, M.G. Lobo: Sipréolico – A Wind Power Prediction System Based on Flexible Combination of Dynamic Models. Application to the Spanish Power System. Proceedings of the World Wind Energy Conference, Berlin (DE), June 2002 [27] Giebel, G. (ed.), J. Badger, P. Louka, G. Kallos, I. Martí Perez, C. Lac, A.-M. Palomares, G. Descombes: Results from mesoscale, microscale and CFD modelling. Deliverable D4.1b of the Anemos project. Available online from Anemos.cma.fr. 101 pages, December 2006 [28] Ed McCarthy: Wind Speed Forecasting in the Central California Wind Resource Area. Paper presented in the EPRI-DOE-NREL Wind Energy Forecasting Meeting March 23, 1998, Burlingame, CA [29] Troen, I, L. Landberg: Short-Term Prediction of Local Wind Conditions. Proceedings of the European Community Wind Energy Conference, Madrid (ES), September10-14, 1990, pp. 76-78, ISBN 0-9510271-8-2 [30] Landberg, L.: Short-term Prediction of Local Wind Conditions. PhD-Thesis, Risø-R-702(EN), Risø National Laboratory, Roskilde, Denmark 1994, ISBN 87-550-1916-1 [31] Troen, I., and E.L. Petersen: European Wind Atlas. Published for the EU Commission DGXII by Risø National Laboratory, Denmark (1998), ISBN 87-550-1482-8 [32] Beyer, H.G., D. Heinemann, H. Mellinghoff, K. Mönnich, and H.-P. Waldl: Forecast of Regional Power Output of Wind Turbines. Proceedings of the European Wind Energy Conference, Nice, France, 1-5 March 1999, pp. 1070-1073, ISBN 1 902916 00 X [33] Mönnich, K.: Vorhersage der Leistungsabgabe netzeinspeisender Windkraftanlagen zur Unterstuetzung der Kraftwerkseinsatzplanung. PhD-thesis, Carl von Ossietzky Universität Oldenburg, 2000. [34] Lange, M., and U. Focken: Physical Approach to Short-Term Wind Power Prediction. Berlin: Springer-Verlag, 2005. [35] Wind Engineering, CARE Special Issue, Vol 23(2), 1999 [36] Kariniotakis, G.N., D. Mayer: An Advanced On-Line Wind Resource Prediction System for the Optimal Management of Wind Parks. Paper presented on the 3rd MED POWER conference 2002, Athens (GR), November 4-6, 2002 [37] Durstewitz, M. C. Ensslin, B. Hahn, M. Hoppe-Kilpper: Annual Evaluation of the Scientific Measurement and Evaluation Programme (WMEP). Kassel, 2001

[38] Ernst, B., K. Rohrig, H. Regber, Dr. P. Schorn: Managing 3000 MW Wind Power in a Transmission System Operation Center. Proceedings of the European Wind Energy Conference, Copenhagen, Denmark, 2-6 June 2001, pp. 890-893, ISBN 3-936338-09-4 [39] Lange, B., K. Rohrig, F. Schlögl, Ü. Cali, R. Mackensen, and L. Adzic: Lessons learnt from the Development of Wind Power Forecast Systems for six European Transmission System Operators. Proceedings of the EWEC, Brussels (BE), 30 March-3 April 2008 (only abstract available) [40] Möhrlen, C., and J.U. Jørgensen: Verification of Ensemble Prediction Systems for a new market: Wind Energy. ECMWF Special Project Interim Reports 1-4, 2004 [41] Jørgensen, J.U., and C. Möhrlen: HONEYMOON - A high resolution numerical wind energy model for on- and offshore forecasting using ensemble predictions. ECMWF Special Project Interim Reports 1-4, 2005 [42] Lang, S., C. Möhrlen, J. Jørgensen, B. Ó Gallachóir, E. McKeogh: Application of a Multi-Scheme Ensemble Prediction System for Wind Power Forecating in Ireland and comparison with validation results from Denmark and Germany. Proceedings of the European Wind Energy Conference and Exhibition EWEC, Athens (GR), 27.2.-2.3. 2006 [43] Lang, S., C. Möhrlen, J. Jørgensen, B. Ó Gallachóir, E. McKeogh: Aggregate Forecasting of Wind Power Generation on the Irish Grid using a Multi-Scheme Ensemble Prediction System. Proceedings, Renewable Energy in Maritime Island Climates, 2nd Conference, Dublin (IE), 26–28 April 2006 [44] Lang, S., C. Möhrlen, J. Jørgensen, B. Ó Gallachóir, E. McKeogh: Forecasting total wind power generation on the Republic of Ireland Grid with a Multi-Scheme Ensemble Prediction System. Proc. Global Wind Energy Conference GWEC, Adelaide (AU), 2006 [45] Bailey, B., M. C. Brower, and J. Zack: Short-Term Wind Forecasting. Proceedings of the European Wind Energy Conference, Nice, France, 1-5 March 1999, pp. 1062-1065, ISBN 1 902916 00 X; see also http://www.truewind.com/. [46] Vidal, J., A. Tortosa, O. Lacave, J. Aymamí, J. Zack, and D. Meade: Validation of a Wind Power Forecast System - The multi-model NWP ensemble strategy. Proceedings of the European Wind Energy Conference, Warsaw (PL), 20-23 April 2010 [47] Zack, J. W.: A new technique for short-term wind energy forecasting: a rapid update cycle with a physics-based atmospheric model. Proceedings of the Global Wind Power Conference and Exhibition, Chicago (US), 28-31 March 2004 [48] González Morales, G.: Sipreólico. Wind power prediction experience. Talk slides accompanied by the paper: Sánchez, I., J. Usaola, O. Ravelo, C. Velasco, J. Domínguez, M.G. Lobo, G. González, F. Soto, B. Díaz-Guerra, M. Alonso: Sipreólico - A wind power prediction system based on flexible combination of dynamic models. Application to the Spanish power system. Proceedings of the First IEA Joint Action Symposium on Wind Forecasting Techniques, Norrköping, Sweden, December 2002, pp.197-214. Published by FOI - Swedish Defence Research Agency. [49] Sánchez, I., J. Usaola, O. Ravelo, C. Velasco, J. Domínguez, M.G. Lobo, G. González, F. Soto: SIPREÓLICO - A Wind Power Prediction System Based on Flexible Combination of Dynamic Models. Application to the Spanish Power System. Poster on the World Wind Energy Conference in Berlin, Germany, June 2002 [50] Sánchez, I.: Short-term prediction of wind energy production. International Journal of Forecasting 22(1), pp. 43-56, 2006 [51] Martí Perez, I., T.S. Nielsen, H. Madsen, J. Navarro, A. Roldán, D. Cabezón, C.G. Barquero: Prediction Models in Complex Terrain. Proceedings of the European Wind Energy Conference, Copenhagen, Denmark, 2- 6 June 2001, pp. 875-878, ISBN 3-936338-09-4 [52] Martí, I., M.J. San Isidro, D. Cabezón, Y. Loureiro, J. Villanueva, E. Cantaro, and I. Pérez: Wind power prediction in complex terrain: from the synoptic scale to the local scale. Proceedings of The Science of making Torque from Wind, Delft (NL), 19-21 April 2004, pp. 316-327 [53] Gow, G.: Short Term Wind Forecasting in the UK. Proceedings of the First IEA Joint Action Symposium on Wind Forecasting Techniques, Norrköping, Sweden, December 2002, pp. 3-10. Published by FOI - Swedish Defence Research Agency.

[54] Westrick, K.: Wind Energy Forecasting in the Pacific Northwestern U.S.. Proceedings of the First IEA Joint Action Symposium on Wind Forecasting Techniques, Norrköping, Sweden, December 2002, pp. 65-74. Published by FOI - Swedish Defence Research Agency. [55] Thompson, G.: Using the Weather Research and Forecasting (WRF) atmospheric model to predict explicitly the potential for icing. Proceedings of the Winterwind conference, Norrköping (SE), 9 Dec 2008 [56] Duncan, T., M. LeBlanc, C. Morgan, and L. Landberg: Understanding Icing Losses and Risk of Ice Trhow at Operating Wind Farms. Proceedings of the Winterwind conference, Norrköping (SE), 9 Dec 2008 [57] Durstewitz, M., J. Dobschinski, Z.Khadiri-Yazami: Wind power forecast accuracy under icing conditions - General approach, practical applications and options for considering effects of wind turbine icing. Proceedings of the Winterwind conference, Norrköping (SE), 9 Dec 2008 [58] Heimo, A.: COST 727 Action Measuring and forecasting atmospheric icing on structures. Proceedings of the Winterwind conference, Norrköping (SE), 9 Dec 2008 [59] Giebel, G.: Wind Power has a Capacity Credit -A Catalogue of 50+ supporting Studies. WindEng EJournal, windeng.net, 2005 [60] Focken, U., M. Lange, K. Mönnich, H.-P. Waldl, H.G. Beyer, A. Luig: Short-term prediction of the aggregated power output of wind farms – a statistical analysis of the reduction of the prediction error by spatial smoothing effects. J. Wind Eng. Ind. Aerodyn. 90(3), pp. 139-249 (March 2002) [61] Boone, A.: Simulation of Short-term Wind Speed Forecast Errors using a Multi-variate ARMA(1,1) Time-series Model. Masters thesis at the Kungliga Tekniska Högskolan, Stockholm, 95 p., 2005 [62] Focken, U., M. Lange, K. Mönnich, H.-P. Waldl, H.G. Beyer, A. Luig: Short-term prediction of the aggregated power output of wind farms – a statistical analysis of the reduction of the prediction error by spatial smoothing effects. J. Wind Eng. Ind. Aerodyn. 90(3), pp. 139-249 (March 2002) [63] Pease, J.: Wind Power Forecasts in the US Context. Talk on the 2nd Workshop on Best Practice in the Use of Short-term Forecasting, Madrid (ES), 28 May 2008 [64] Pease, J.: Critical Short-term Forecasting Needs for Large and Unscheduled Wind Energy on the BPA System. Talk on the 3rd Workshop on Best Practice in the Use of Short-term Forecasting, Bremen (DE), 13 October 2009 [65] Focken, U.: Experiences with Extreme Event Warning and Ramp Forecasting for US Wind Farms. Talk on the 4th Workshop on Best Practice in the Use of Short-term Prediction of Wind Power, Quebec City (CA), 16 October 2010 [66] Wilson, N., R. Pyle, D. Atallah, and K. Parks: State-of-the-Art Wind Energy Ramp Event Forecasting Using Atmospheric Observations. Proceedings of the EWEA Annual Event, Brussels (BE), 14-17 March 2011 [67] Myers, W., and S. Linden: A turbine hub height wind speed consensus forecasting system. 91st American Meteorological Society Annual Meeting, Seattle (US), 22-27 January 2011 [68] Haupt, S.E., G. Wiener, Y. Liu, B. Myers, J. Sun, D. Johnson, and W. Mahoney: A Wind Power Forecasting System to Optimize Power Integration. Proceedings of the ASME 5th International Conference on Energy Sustainability, Washington (US), 7-10 August 2011 (read as pre-conference draft) [69] Davy, R.J., M.J. Woods, C.J. Russell and P.A. Coppin: Statisical Downscaling of Wind Variability from Meteorological Fields. Boundary Layer Meteorology 135(1), pp. 161-175, 2010 [70] Vincent, C.L., P. Pinson and G. Giebel: Wind fluctuation over the North Sea. International Journal of Climatology. 2010 [71] Vincent, C.L., G. Giebel, C.J. Russel, M.J. Woods, R.J. Davy: Documentation of the severity index for wind variability and its application to wind power forecasting. Deliverable report D4.7 for the SafeWind project. [72] Bremnes, J.B.: Probablilistic wind power forecasts by means of a statistical model. Proceedings of the First IEA Joint Action Symposium on Wind Forecasting Techniques, Norrköping, Sweden, December 2002, pp. 103- 114. Published by FOI - Swedish Defence Research Agency. [73] Bremnes, J.B.: Probabilistic Wind Power Forecasts Using Local Quantile Regression. Wind Energy 7(1), pp. 47-54, 2004 [74] Pinson, P., G. Kariniotakis: On-line Assessment of Prediction Risk for Wind Power Production Forecasts. Proceedings of the European Wind Energy Conference and Exhibition, Madrid, Spain, 16-19 June 2003.

[75] Pinson, P., and G. Kariniotakis: On-line Adaptation of Confidence Intervals based on Weather Stability for Wind Power Forecasting. Proceedings of the Global Wind Power Conference and Exhibition, Chicago (US), 28-31 March 2004 [76] Nielsen, H.Aa., T.S. Nielsen, H. Madsen, J. Badger, G. Giebel, L. Landberg, H. Feddersen, K. Sattler: Wind Power Ensemble Forecasting. Paper presented on the Global Wind Power Conference and Exhibition, Chicago (US), 28-31 March 2004 [77] Taylor, J.W., P.E. McSharry, and R. Buizza: Wind Power Density Forecasting Using Ensemble Predictions and Time Series Models. IEEE Transactions on Energy Conversion 24, pp. 775-782, 2009 [78] Pierre Pinson, Henrik Madsen: Ensemble-based probabilistic forecasting at Horns Rev. Wind Energy 12(2), pp.137-155, 2009 [79] Carl Hilger, Eltra, at the Fuel and Energy Technical Association Conference on “Challenges from the rapid expansion of wind power” on 3rd April 2005 (oral statement) [80] Parks, K.: Xcel Energy/NCAR Wind Energy Forecasting System. Talk on the UWIG Forecasting Workshop, Albany (US), 23-24 February 2011

3.1.9 MEASUREMENTS9

Some of the meteorological instruments that are frequently used in wind energy studies are described in the following:.

Cup-anemometers and wind-vanes

The classical way to measure wind speed is by use of a cup-anemometer. This measuring device is widely used because of robustness and durability combined with a good accuracy of the wind speed measurements, being typically 1% when calibrated according to the MEASNET standard.

It consists of typical 3 cups mounted on a vertical axis. The drag on the cups arising from the wind is larger for flows towards the open side of the cups than the closed side, and the induced pressure difference makes the cups rotate. The speed of rotation is near proportional to the wind speed. A disadvantage of the cup anemometer is the threshold velocity, being the wind speed at which the cups start to rotate; it is typically 0.3 ms-1. Because a cup-anemometer reacts faster to an increase in the wind speed than to a decrease, it will in a gusty wind climate measure a higher wind speed than actually present – the so-called over-speeding; the effect depends on the design of the cup-anemometer [1]. More details on the cup-anemometer can be found in Gryning (1982).

Wind direction is traditionally measured with a wind vane, being a vertical blade that is attached to a vertical axis. The wind will turn the blade to align with the wind direction, the behavior of the wind vane in a fluctuation wind field is discussed in [2] and [3].

Wind lidars

Light emitted into the air interfere with it. This effect is used in a wind lidar, where the Doppler shift of the backscattered laser light is used to determine the wind speed along the line of sight. Exploiting recent developments in optical fiber technology (from the telecommunication sector) to construct the interferometer that constitutes the core of a lidar, has made it possible to develop of rather small and transportable wind lidars.

9 Lead author: Sven-Erik Gryning

Range resolves lidar’s transmit pulses that are scattered by particles in the air and a very small part is returned to the receiver. The distance to the measurement volume can be determined from the time of flight of the light pulses. Comparing the lidar resolved wind speed measurements with cup-anemometer wind speeds reveals a very good agreement. Presently lidar based profiles of wind speed are feasible up to 2 km with a vertical range of 50 meters but the technique is continuously developing.

Alternatively, the wind speed profiles can be estimated from beam focusing. The lidar emits a continuous wave beam that is focused at specified heights. The profile is obtained by scanning a cone. A drawback of this system is that the probe length increases with height, being proportional to approximately the square of the height. Therefore the system provides an excellent vertical resolution near the ground, say up to typically 100 meters.

SCADA

Each wind turbine is equipped with a system that controls and logs the conditions for the wind turbine. The system is called Supervisory Control and data Acquisition (SCADA). These system logs a large number of vital parameters for the wind turbine, including the wind speed and direction. The wind speed is measured at the nacelle and therefore influenced by the wind turbine itself, but the nacelle wind speed has been used in many investigations as a substitute for the undisturbed wind. The use of SCADA wind speeds is still a matter of discussion in the literature and no general conclusions seems yet to have been reached upon its proper use and for what specific conditions it is preferred.

References to §3.1.10:

[1] Busch, N. E. and Kristensen, L. (1976). Cup anemometer over-speeding. J. Appl. Meteorol., 15, 1328-1322. [2] Larsen, S. E. and Busch, N. E. (1974). Hot-wire measurements in the atmosphere. Part 1: Calibration and response characteristics. DISA Information 16, 15-34.

[3] Gryning, S. E. Elevated Source SF6-Tracer Dispersion Experiments in the Copenhagen Area. Risø-R-446. 187 pp.

3.2 SOLAR ENERGY10

3.2.1 OVERVIEW

Solar power, both from photovoltaic (PV) and solar thermal power plants, are – together with wind power – the major contributor to a future energy system which is largely based on renewable energies. However, in the electricity sector both sources introduce new challenges to the system‘s operation due to their strong intermittent nature. Power generation from solar and wind energy systems shows fundamentally different characteristics than from conventional energy sources. The availability of solar and wind energy is largely determined by the prevailing weather conditions and is therefore highly variable.

Any adaptation within the electric power system to this new constellation – may it be intelligent power plant scheduling, demand side management, or on the long-term the re-structuring of the grid topology and the introduction of storage capacities – will require very detailed information on the expected power production from these sources on various temporal and spatial scales. Reliable forecasts of solar power production therefore are an essential factor for an efficient integration of large amounts of solar power into the electric supply system. The first prediction systems for photovoltaic power became operational recently [1][2][3][4] and already contribute to an increase in the economic value of the PV energy produced. As the benefit of solar power forecasts for grid integration is directly related to their accuracy, increasing effort is currently spent on research on solar irradiance forecasting as a basis for corresponding solar power forecasts.

Depending on the application and its corresponding time scale, different forecasting approaches have been introduced. Time series models using on-site irradiance measurements are adequate for the very short term time scale from minutes up to a few hours. Intra-hour forecasts of clouds and irradiance with a high spatial and temporal resolution may be obtained from ground-based sky imagers. Forecasts based on cloud motion vectors from satellite images show a good performance for a temporal range of 30 minutes to 6 hours. Grid integration of PV power mainly requires forecasts up to 2 days ahead or even beyond. These forecasts are based on numerical weather prediction (NWP) models.

This chapter provides a brief introduction to the various applications of solar power forecasting and to the models used for irradiance and power prediction for the different time domains, as well as presents exemplary results for selected prediction systems.

3.2.2 APPLICATIONS OF IRRADIANCE AND SOLAR POWER FORECASTS

3.2.2.1 GRID INTEGRATION OF PV POWER

The most important application of PV power forecasts is to support a cost-effective integration of large amounts of solar power into the electricity supply system. The required forecast horizon is partly defined by technical constraints, e.g., the mix of power plants in energy systems with their specific characteristics, e.g., start-up times. From a market perspective, the trade of PV power on the so-called day-ahead market is an important application for PV power forecasting. Here bids for the next day have to be placed typically around noon, announcing the supply of or the request for electric power in dependence on the daytime. In both cases, power forecasts with a forecast horizon

10 Lead author: Detlev Heinemann

up to 2 days ahead are of particular relevance for grid integration purposes. Additionally, energy markets offer the possibility of intraday trading, in which the day-ahead schedule is updated during the day and more precise short- term forecast information can be used. In this case, the time horizon is in the range of some hours and a high-quality forecast may save costly balancing power to be supplied from conventional sources.

The need for forecast information on the expected renewable power production strongly increases with the capacity of installed power. Wind power prediction systems already have shown their strong economic impact and benefit for the integration of wind energy into the electricity grid (see Chapter 3.1). With installed PV capacities reaching the same magnitude as for wind installations, the prediction of solar power gain more and more economic relevance. For example, in Germany, about 25 GW of PV power were installed at the end of 2011, reaching a share of 3% of the total electricity supply. On sunny summer days around noon, solar energy contributes up to 40% of the overall electricity demand. Due to the partially high correlation of load and PV power on sunny days, PV power shows a significant peak shaving capability which is even more pronounced in hot and sunny regions, where a considerable amount of electricity is used for cooling (e.g., see [5]).

As a consequence of this new and rapidly evolving situation on the energy market, various operational PV power prediction systems have been introduced recently [1][2][3][4], and respective services are requested, for example, by grid operators. The use of these forecasts and the corresponding spatial and temporal scales depend on the specific regulatory framework applied in the respective countries. In Germany, according to the current feed-in law, PV system operators may feed in their complete electricity production with priority to conventional electricity for a fixed price. Hence, grid operators are in charge of balancing the fluctuating input for the corresponding control areas, and regional forecasts are required. In Spain solar plant operators are offered two choices [6]. A fixed tariff model similar to the German feed-in tariff model offers system operators a guaranteed right to feed in all PV power and they receive a fixed price. In addition, operators of plants with an installed power of more than 1 MW are allowed to directly participate in the electricity market, placing bids on the day-ahead and intraday market. By choosing this model, solar plant operators are obliged to deliver power according to the schedule specified in the contract. Otherwise they are charged a penalty, depending on the deviation of delivered from scheduled power. Using the premium tariff model, an additional premium per kWh is paid by the Spanish National Energy Commission. When participating in the premium tariff model, plant operators will need site-specific forecasts of the produced power for the current and next day. In Japan, PV power has to be delivered to the grid according to a day-ahead schedule, and schedule violations are associated with an additional fine. In [7], the addition of a storage system and its impact on controllability and adjustability of PV power, is investigated in combination with day-ahead and current-day forecasts of solar irradiance. The end-use accuracy of power forecasts for grid integration is subject of two case studies performed in the United States [5]. The ability of PV power to contribute at critical load demand times is highlighted. Forecasting the peak load reduction by PV power has shown to be very beneficial for an adjustment of the predicted load demand for conventional generation. Simulations of the capability of PV to reduce the peak load for different scenarios of grid penetration of PV have shown the high value of solar power forecasts for his application and that they can contribute to an effective use of solar resources in the power grid.

3.2.2.2 CONCENTRATING SOLAR POWER (CSP)

Concentrating solar thermal power plants (STPP) use various mirror configurations to convert solar energy into high- temperature heat which in a second stage is used to conventionally produce electric power. Direct conversion of solar energy to electricity is performed in concentrating photovoltaic (CPV) devices which use optics such as lenses for concentration. Mostly the term CSP refers to solar thermal systems only. Both technologies rely on direct solar irradiance instead of global irradiance.

Regarding forecasting of direct normal irradiance (DNI), this adds an additional problem, since DNI is usually not a prognostic variable in NWP models. Consequently, additional post-processing is needed to derive DNI from the model information. In [13] a model is proposed based on the combined use of information provided by a NWP

model, an air-quality model and remote sensing retrievals. The model substantially relies on forecasts of the aerosol load [14]. Evaluation for sites in Europe and Northern Africa result in relative RMSE values of around 20% for the DNI forecasts under clear sky conditions. A method to derive DNI forecasts from the WRF (Weather and Research Forecasting) model [39], a NWP model, and satellite retrievals was presented in [58]. The methodology, fully operational, was evaluated in southern Spain [48] and showed that the WRF model presents considerable skill in both, GHI (global horizontal) and DNI forecasting, with relative RMSE values about 5% for the GHI and around 15% for the DNI under clear-sky conditions. However, for other sky conditions the forecast quality considerably diminishes. Particularly, under broken cloud conditions relative RMSE values approach 45%. Overall, the WRF model was found to provide considerable bias.

Usually, NWP account for the aerosol extinction from climatological values (mostly monthly-based). However, in a clear day, the solar extinction by aerosols is a key factor which may reach 20% of the DNI [58] thus being a potential source of large errors. Therefore, an appropriate aerosol representation in NWP models is crucial for accurate DNI forecasting. Currently, satellite retrievals, as those from MODIS, or wide ground networks, as AERONET, offer valuable information about the near-real-time aerosol content of the atmosphere from around a decade ago. Recently, the WRF code has been extended to separately calculate direct and diffuse irradiances and to be able to ingest aerosol estimates from MODIS imagery [59].

In the framework of the international MACC project (Monitoring atmospheric composition & climate) aerosol forecasts are developed using a bulk aerosol scheme with a small number of representative aerosol variables [60]. For data assimilation MODIS aerosol optical depth measurements are used. Currently, this scheme is incorporated within ECMWF’s Integrated Forecasting System (IFS).

3.2.2.3 STAND-ALONE SYSTEMS AND SMALL NETWORKS

The performance of small electric networks including PV as well as of stand-alone PV systems can be improved when solar forecast information is used. Several studies have recently reported different applications of forecasting in this domain. A power forecasting system designed to optimize the scheduling of a small energy network including PV is described in [8]. Electric power and heat in these networks may be controlled with advanced communication networks and power predictions for a forecast horizon up to 24 hours are used in corresponding operating strategies, especially for optimizing the energy flow between the electric and thermal system with fuel cells operating at the interface. The influence of weather disturbances on the stability of the Kythnos island micro-grid power system has been studied in [9]. While looking at the system stability as a function of PV penetration they demonstrated that the stability may be improved when information on cloud cover approaching the island is available 15 minutes in advance to allow for start-up of power backup or the disconnection of less critical loads. The authors recommend satellite-based nowcasting and short-term forecasting to obtain the required information on cloud motion. The use of irradiance forecasts for a stand-alone PV system is presented in [10]. A forecast-based control method is used for a PV-diesel hybrid generation system on a ship. Irradiance forecasts for the current day with hourly resolution are used in the operating strategy to maximize the diesel engine‘s efficiency and to minimize contributions to and from the battery. This finally allows for a reduction of the battery size.

3.2.2.4 OTHER APPLICATIONS

Finally, it shall be mentioned that there are numerous applications of irradiance forecasting in the energy sector not related to grid integration issues or even not to electricity at all. Examples include the use of weather and solar irradiance forecasts for the control of heating, ventilating, and cooling of buildings [16]; the use of irradiance forecasts to improve the management of district heating grids that integrate solar thermal water heating; load forecasting [16]; and the use of forecasts in agriculture, for example, for crop harvesting.

3.2.3 MODELS FOR THE PREDICTION OF SOLAR IRRADIANCE AND PV POWER

3.2.3.1 TYPICAL OUTLINE OF A PV POWER PREDICTION SYSTEM

Power prediction of PV systems usually involves several modeling steps in order to obtain the required forecast information from different kinds of input data. A typical model chain of a PV power forecasting system comprises the following basic steps (Figure 7): • Forecast of site-specific global horizontal irradiance • Forecast of solar irradiance on the module plane • Forecast of PV power. Regional forecasts need an additional step for upscaling: • Forecast of regional power production.

These steps may involve physical or statistical models or a combination of both. Not all approaches for PV power prediction necessarily include all modeling steps explicitly. Several steps may be combined within statistical models, for example, relating power output directly to input variables like measured power of previous time steps or forecast variables of NWP systems.

Forecasting of global horizontal irradiance is the first and most essential step in almost any PV power prediction system. Depending on the forecast horizon, different input data and models may be used.

• In the very short-term time scale from minutes to a few hours, on-site measured irradiance data in combination with time series models are appropriate. Examples of direct time series models are Kalman filtering, autoregressive (AR), and autoregressive moving average (ARMA) models. Artificial neural networks (ANNs) are used as well within this category. • In short-term irradiance forecasting, information on the motion of clouds which largely determine solar surface irradiance may be used. Forecasts based on satellite images show a good performance for up to 6 hours ahead. From subsequent images information on cloud motion can be extracted and extrapolated to the next few hours. For the sub-hourly time scale, cloud information from ground-based sky imagers may be used to derive irradiance forecasts with much higher spatial and temporal resolution compared with satellite data. Forecast horizons are limited here through the spatial extension of the monitored cloud scenes and corresponding cloud velocities. • From about 4–6 h onward, forecasts based on NWP models typically outperform the satellite-based forecasts. The weather service’s global NWP models describe the complete Earth with a comparatively coarse spatial and temporal resolution. Meso-scale models allow for calculations on a finer grid covering selected regions. Boundary and initial conditions are then provided by a global model. However, recent improvements in the resolution of the global models more and more make this difference less critical. Some weather services, for example, the European Centre for Medium-Range Weather Forecasts (ECMWF), directly provide surface solar irradiance as model output. This allows for site-specific irradiance forecasts with the required temporal resolution produced by downscaling and interpolation techniques. Statistical models may be applied to derive surface solar irradiance from available NWP output variables and to adjust irradiance forecasts to ground- measured or satellite-derived irradiance data.

From horizontal irradiance, the irradiance on the plane of the PV modules has to be calculated next. Different installation types have to be considered: • For systems with a fixed orientation, forecasts of global horizontal irradiance have to be converted according to the specific orientation of the modules. Models for this task require information on tilt and azimuth of the PV system. • For one- and two-axis tracking systems, these models have to be combined with respective information on the tracking algorithm.

• Concentrating PV systems require forecasts on direct normal irradiance. The procedure is the same as with any concentrating system, e.g., solar thermal power plants.

The PV power forecast then is obtained by feeding the irradiance forecast into a PV simulation model. Generally, two models are used in this step: One for the calculation of the direct current (DC) power output and another for modeling the inverter characteristics. Both models are widely available in the PV sector with various degrees of complexity. For PV power prediction, rather simple models show a sufficient accuracy, which is higher than the accuracy of the irradiance forecast. Additional input data are module temperature, which can be inferred from available temperature forecasts, and the characteristics of the PV module (nominal power etc.), usually taken from the module data sheets.

In the final stage towards an optimized power forecast for a single PV system, the power forecast may be adapted to measured power data by statistical post-processing techniques. Self-calibrating recursive models are most beneficial if measured data are available online. Off-line data are successfully used as well for model calibration.

Figure 8: Typical model chain for PV power prediction

PV power prediction for utility applications usually addresses the cumulative PV power generation for a larger area rather than for a single site. This is achieved by up-scaling from a representative set of single PV systems to the regional PV power production. This approach leads to almost no loss in accuracy when compared to the addition of the complete set of site-specific forecasts if the representative set properly represents the regional distribution of installed power and installation type of the systems. In addition to the power prediction, a specification of the expected uncertainty of the predicted value is important for an optimized application. This uncertainty information provides the basis, for example, for an assessment of the risk associated with decisions based on the forecasts or for an estimation of the necessary reserve power determined by the largest forecast errors [12]. As the correlation of forecast errors rapidly decreases with increasing distance between the systems, the uncertainty associated with regional power prediction is generally much smaller than for single PV systems.

In the following, the different stages of a complete PV power prediction scheme are described.

3.2.3.2 IRRADIANCE FORECASTING

Solar power production is essentially determined by the incoming solar irradiance, which therefore has to be in the center of any power forecasting scheme. As outlined in the previous chapter, irradiance forecasting approaches mainly differ according to the time scale of the application. These approaches are briefly described below.

3.2.3.2.1 TIME SERIES MODELING

Time series models provide solar irradiance forecasts using only measured values of solar irradiance as input. Further measurement values related to solar irradiance, for example, cloud cover, may be included as well. Time series models make use of the high autocorrelation for short time lags in time series of solar irradiance and cloud cover. For very short-term time scales, typically up to 1 or 2 h ahead, time series forecasts based on accurate on-site measurements will be advantageous. Generally, two principal time series approaches may be distinguished: The statistical or direct time series approach, and the learning or artificial intelligence (AI) approach.

In the statistical approach, relations between input variables to the model (predictors), and the quantity to be predicted (predictand), are derived from statistical analysis. An early approach in direct time series irradiance forecasting based on autoregressive integrated moving average (ARIMA) models has been proposed in [16]. Since then, several studies with respect to direct time series modeling have been performed. In [17] different time series models are compared, and in [18] the use of a simpler autoregressive (AR) model to directly predict PV power in comparison with other models is investigated.

Artificial neural networks (ANNs) may overcome the limitations of conventional linear approaches and solving complex and nonlinear problems that are difficult to model analytically. The relation between the desired output and input data is learned using data of a training set. In solar irradiance forecasting, the output of the ANN is the solar irradiance at a certain time and the inputs are irradiance values or related meteorological parameters at previous time steps. Irradiance forecasting approaches based on ANN and other AI techniques have been proposed by several research groups [20][21][22][23][24][25]. Overviews are given in [19] and [21].

Both time series approaches may include not only on-site measured data but also additional input from NWP models. This allows for an extension of the forecast horizon from some hours to some days. Also, this is an example of the widely used approach to combine various techniques in favor of an increase in accuracy or an extension of the time horizon.

3.2.3.2.2 CLOUD MOTION VECTORS

With increasing forecast horizon, time series models are more and more unable to represent the development of clouds which is the major influence on the temporal and spatial variability of irradiance. For forecast horizons up to some hours, the temporal change of cloud structures is mainly governed by horizontal advection thus leaving the shape of clouds in the relevant spatial scale rather stable. Only when strong thermal processes with convection etc. occur the forecast horizon may reduce. Any technique which is able to detect this horizontal cloud motion in sufficient detail should provide valuable information for a cloud motion forecast in the corresponding time scales. Currently, satellite images and ground-based sky imagers are common sources for this data.

Cloud motion vectors from satellite images

Geostationary satellites with their high temporal and spatial resolution offer the potential to derive the required information on cloud motion. Cloud motion vectors are commonly used in operational weather forecasting to describe wind fields in upper heights [32]. Approaches to forecast solar irradiance based on Meteosat satellite images as a basis for a PV power forecast are described in [27][28][29]. In [30] and [31] irradiance forecasts based on the

Geostationary Operational Environmental Satellite (GOES) and the Geosynchronous Meteorological Satellite (GMS) with an approach following [29] are reported.

A common scheme to derive solar irradiance forecasts from cloud motion vectors as described in [29] consists of the following (Figure 8): • As a measure of cloudiness, cloud index images according to the Heliosat method [33], a semi-empirical method to derive solar irradiance from satellite data, are calculated from the satellite data. • Motion vector fields are calculated from consecutive cloud index images. • The future cloud situation is estimated by the extrapolation of motion, i.e., by applying the calculated motion vector field to the current image. • A smoothing filter is applied to the predicted cloud index image in order to eliminate randomly varying small- scale structures that are not predictable. Filtering this ‘noise’ considerably improves the forecast accuracy. • Surface solar irradiance is derived from the smoothed forecast cloud index images using the Heliosat method.

Cloud motion vectors from ground-based sky imagers

Information on cloud motion as a basis for short-term forecasting may also be derived from ground-based sky imagers [34]. Compared with satellite data, ground-based sky imagers offer a much higher spatial and temporal resolution, including the potential for capturing sudden changes in the irradiance – often referred to as ramps – on a temporal scale of less than 1 minute. The maximum possible forecast horizon strongly depends on the cloud condition and is limited by the time the monitored cloud scene has passed the location or area of interest, which depends on the spatial extension of the monitored cloud scenes in combination with its velocity. In [34] forecasts up to 5 minutes ahead were evaluated for four partly cloudy days. An estimation of a maximum possible extension of the forecast horizon in dependence on the cloud scene resulted in values ranging from 5 to 25 minutes. The forecasting procedure involves similar steps as described for satellite-based forecasts.

Figure 9: Short-term forecasting scheme using cloud index images. 45

3.2.3.2.3 NWP-BASED IRRADIANCE FORECASTS

Numerical weather prediction (NWP) models are operationally used to forecast the state of the atmosphere up to 15 days ahead. The atmospheric dynamics, i.e., the temporal changes, are modeled by numerically solving the basic differential equations that describe the physical laws governing the weather. Starting from initial conditions that are derived from worldwide observations, the future state of the atmosphere is calculated in a first step with global NWP models, which are currently run by about 15 weather services. Examples are the Global Forecast System (GFS) run by the US National Oceanic and Atmospheric Administration (NOAA) and the Integrated Forecast System (IFS) operated at the ECMWF. Global models usually have a coarse resolution and do not allow for a detailed mapping of small-scale features, although resolution has increased rapidly during the last years (depending on the model) and is nowadays in the range of 16–50 km. In the next step, different concepts may be applied to account for local effects and to derive improved site-specific forecasts. One possibility is the downscaling by meso-scale models. Mesoscale models cover only a part of the Earth but can be operated with a higher spatial resolution. They are routinely run by national weather services and private weather companies. Also, post-processing methods, for example, model output statistics (MOS), may be applied to model local effects. In addition, they allow for the correction of systematic deviations in dependence on different meteorological parameters and for modeling of the irradiance if irradiance is not provided as output parameter of an NWP model. Post-processing may be applied directly to the output of a global model and likewise also to regional model output.

Irradiance forecasts of the ECWMF global model

The ECMWF provides weather forecasts up to 15 days ahead. It is described here because they have proven their high quality as a basis for both wind and solar power forecasts and surface solar irradiance and different cloud parameters are direct model output instead of diagnostically calculated from other parameters as this is done in many other models. The model performance is constantly increasing and – from the solar power forecasting perspective – the increasing grid resolution and the improvements with respect to radiation and cloud schemes are worth to be mentioned.

The model resolution has increased from approximately 200 km x 200 km horizontal resolution and 16 vertical levels in its early days in 1985 to 16 km x 16 km and 91 vertical levels in 2010. The temporal resolution of the forecasts is 3 h for the first 3 forecast days that are most relevant for solar power prediction.

Modeling of radiation transfer is among the most computational expensive parts in NWP models. It requires the characterization of all interactive processes in the atmosphere, i.e., scattering and absorption, in a variety of different spectral domains. These processes are highly related to the occurrence of clouds in the atmosphere, which are extremely variable in space and the representation of their optical and microphysical properties in atmospheric models with coarse resolution is one of the major challenges in atmospheric modeling. The currently implemented radiation scheme including radiation parameterizations and aerosol description can be seen as today’s state-of-the-art for these processes in NWP modeling [37]. It includes a short-wave radiation scheme with 14 spectral intervals and revised cloud optical properties.

Irradiance forecasting with the meso-scale models MM5 and WRF

Meso-scale models are restricted to limited areas, but generally solve the governing equations on a much finer grid which makes them capable of resolving smaller atmospheric phenomena as land-sea breezes, thunderstorms, and topographically forced wind flow. They have been developed for special applications (local forecasting, extreme situations, complex topographies ...) but are nowadays widely used tools for everyday weather forecasting. The spatial and temporal dynamics of surface solar irradiance is heavily influenced by small-scale features like broken cloud fields and heterogeneous surface characteristics. This is why predictions of solar irradiance with meso-scale models are potentially advantageous.

Only two models which have seen a wide application in energy meteorology, MM5 [38] and WRF [39], shall be mentioned here. Both are developed as open source models in a collaborative effort of several institutes lead by the National Center for Atmospheric Research (NCAR) in the US. They are non-hydrostatic meso-scale models using terrain-following coordinates with multiple nesting capabilities. WRF can be seen as a follow-up model to MM5. Both, WRF and MM5 offer various parameterizations for the different physical processes which allows its configuration for specific conditions for the region of interest with irradiance as designated output parameter. Also, they are able to integrate local measurements, for example, aerosols. As limited-area models, meso-scale models require input from global NWP models for initialization and boundary conditions. Usually, the input data chosen strongly influences the performance of a meso-scale model. To achieve the intended high spatial resolution in a meso-scale model with reasonable computing time, a stepwise increase of the spatial resolution is achieved by a nesting procedure as shown in Figure 9. MM5 and WRF offer different long- and shortwave schemes for radiation transfer and several parameterizations for cloud physical processes. However, different from global models, clouds are represented as horizontally homogenous within the model grid boxes and no sub-grid variability is considered.

Studies on the performance of MM5 with respect to solar irradiance forecasting in the context of model development are reported in [41][42][43]. Generally, a satisfying simulation of cloud and radiation fields is achieved, whereas the representation of small-scale cloud structure (different layers, sub-grid clouds) and the effect of aerosols are considered as weak points. Various configurations of the cumulus, moisture, and planetary boundary layer parameterizations and the influence of the topographic parameterization with respect to irradiance calculations have been investigated in [44] and [45]. Further studies using WRF for irradiance and solar power prediction have been presented in [46][47][48].

Figure 10: Nested domains of the WRF model used in [40]. The outermost domain (green), middle domain (red), and the inner study area (blue) have spatial resolutions of 27 km×27 km, 9 km×9 km, and 3 km×3 km, respectively.

3.2.3.2.4 POST-PROCESSING OF NWP MODEL OUTPUT

The output of NWP models is frequently refined to account for detailed local weather features which are generally not resolved by these models. In addition, systematic deviations for certain weather situations are observed in both global and meso-scale model predictions. To correct these uncertainties, statistical or other post-processing methods

have been introduced, and most irradiance forecasting approaches make use of them. In particular, post-processing methods • reduce systematic forecast errors (correction of systematic deviations), • account for local effects (e.g., topography), • account for the influence of selected variables in more detail (e.g., aerosols), • derive parameters that are not directly provided by the NWP models (e.g., solar surface irradiance if not a standard output parameter), and • combine the output of different models.

Model output statistics

Model output statistics (MOS) is an established technique based on regression equations and widely used to refine the output of NWP models to account for local variations in surface weather [50]. This is done by using, e.g., additional surface observations and climatology for specific locations. Observations of the designated forecast variable – the predictand – are related to model forecast variables – the predictors – with a statistical approach. The set of predictors may be extended by including any relevant information, for example, prior observations and climatological values. Most important are high-quality measurements from local surface stations. In case of solar irradiance, also satellite-derived values may be used.

MOS was used in the first published effort to forecast solar radiation [51] where daily forecasts for one and two days in advance were produced using predictors including cloud information. More recent attempts use ECMWF output, statistically derived predictors, and surface or satellite observations of irradiance [1]. A comparison of irradiance forecasts using this MOS scheme with WRF forecasts and other approaches is given in [40] and [49]. Sky cover forecasts of the US National Digital Forecast Database (NDFD) which are generated as a combination of US national model output, meso-scale model runs, and human input, are used with good results in [54] for solar irradiance forecasts. Here, a typical feature of MOS is used, which is the inference of a parameter that otherwise would not be available.

Systematic deviations of NWP output variables often depend on the meteorological situation. In [35], a bias correction in dependence on the predicted cloud situation for the application to ECMWF irradiance forecasts is introduced. The original forecasts show a considerable overestimation of irradiance for intermediate cloud cover. The proposed approach for bias correction has been adapted and evaluated also for other NWP models and different climates in two recent studies [55]. In [56] an analysis of irradiance forecasts of three different NWP models (GFS, North American Model (NAM), and ECMWF) for stations in the continental United States is provided. For all the three models, forecast accuracy could be improved by applying the weather-dependent bias correction.

Any statistical approach relating observed variables to NWP output is by definition similar to the concept of MOS. In particular, Artificial Neural Networks (ANN) have also been used to improve NWP output with respect to irradiance prediction [26][52][53].

Temporal interpolation

Global model forecasts are provided with a temporal resolution of 3–6 h. Many applications in the energy sector, however, need forecasts of solar power at least on an hourly basis. Using various interpolation techniques, hourly forecasts are derived from global NWP output. In [35], the combination of irradiance forecast data with a clear-sky model to account for the typical diurnal course of irradiance is proposed.

Spatial averaging

Forecast accuracy can be improved by applying a smoothing filter or spatial averaging. This reduces fluctuations of forecast values in variable cloud situations where this is favorable, but preserves the quality of the original forecasts

in homogeneous clear-sky and overcast situations. For ECMWF forecasts with an original spatial resolution of 25km×25km and temporal resolution of 3 h, best results are achieved for average values of 4×4 grid points corresponding to a region of 100km×100km [35]. The impact of spatial averaging is much stronger for meso-scale or multi-scale model output with hourly values and a finer grid resolution.

Physical post-processing approaches

Physical post-processing using radiation transfer calculation is another option to improve NWP output. Benefits are the consideration of additional parameters that are not handled in detail in NWP models, like aerosols, and the potential to provide direct normal irradiance (DNI) as well, which is relevant for all concentrating solar power plants. A forecasting system for global and direct irradiance including aerosol forecasts derived with a chemical transport model is proposed and evaluated in [13]. Irradiance calculations are performed with the radiative transfer library libRadtran [15]. Model input are aerosol forecasts, cloud cover and water vapor forecasts with MM5, and satellite- derived surface albedo and ozone values. In [48], the authors analyze the reliability of 3-day-ahead hourly global irradiance and DNI forecasts in Andalusia (South of Spain) based on the WRF meso-scale model. The DNI forecasts are calculated on the basis of WRF model output and satellite retrievals with a physical post-processing procedure based on radiation transfer calculations. A semi-physical post-processing procedure for topographic downscaling of solar irradiance forecasts in mountainous regions to account for geometric shading and reflection, is proposed in [58]. A digital elevation model with a spatial resolution of 90m×90m ist used to provide the topographic information.

Human interpretation of NWP output

A traditional method to obtain improved local forecasts from NWP model output is to participate from the expert knowledge of a human forecaster [11]. Here, cloud cover forecasts of meteorologists are combined with a with a clear-sky model. Especially in difficult forecast situation, for example fog, this offers potential for further improvements of the irradiance forecast.

3.2.3.3 PV POWER FORECASTING

Forecasts of global irradiance as described in the previous section provide the basis for most PV power prediction schemes. Different approaches to infer aggregated electric power output from irradiance forecasts have been proposed: • Physical approaches explicitly model the processes determining the conversion of solar irradiance to electricity. This includes conversion of irradiance from the horizontal to the array plane and PV simulation (Figure 7). Temperature information is useful as an additional forecast parameter influencing PV power generation. • Statistical or learning algorithms do not model the physical processes directly but try to establish the relation between power output and irradiance forecasts on the basis of historical data sets. Input data to statistical approaches are not limited to irradiance forecasts, but may be, for example, prior PV power output observations. • Combined or hybrid approaches apply physical models first and then correct the results with a statistical model.

Explicit PV power prediction approaches are presented, for example, in [1][4][25][36]. Examples of purely statistical methods are given in [61], where PV power forecasts are derived from prior observations of PV power, different irradiance components, and temperature, and in [18], where PV power is derived from prior observations and NWP forecast variables using AR models with exogenous input. A direct comparison of an explicit and a statistical model based on NWP input is shown in [8], with better results for the statistical approach. In [36], a hybrid method is proposed, involving statistical adjustments at different stages of the modeling chain.

Power prediction for utility applications usually requires up-scaling to regional power as a final step. Only few studies have addressed this so far. A preparatory study is presented in [1] and further work is described in [4].

3.2.3.3.1 IRRADIANCE ON THE MODULE PLANE

The first stage in any PV power calculation from irradiance data is the calculation of the proper irradiance values for the given module plane, usually from available horizontal global irradiance. Systems with fixed tilt angles, tracked systems, and concentrating systems have to be treated differently at this stage.

A large number of empirical models for this conversion have been developed that may be used as well for PV power prediction purposes. Mostly, the models try to decompose the global irradiance into its direct beam, sky diffuse, and ground-reflected radiation, which then are converted separately to the tilted plane (Figure 10). Although the basic principle is straight forward, the conversion implies the use of two empirical relationships with its inherent uncertainties. First, it needs the separation of horizontal beam and diffuse irradiance given only the global irradiance and geometric information. A common approach is to describe this via an empirical function of the clearness index (i.e., ration of global irradiance and extraterrestrial irradiance) and – in some models – of a cloud variability parameter [62]. Second, any geometric conversion of diffuse irradiance needs information about the directional distribution of radiance over the sky, which especially for clear skies is strongly anisotropic. Empirical models which take this anisotropic effects for modeling the diffuse component into account, are e.g. [63] and [64].

Figure 11: Conversion steps for estimating global irradiance on tilted surfaces from global horizontal irradiance. The components framed in red are empirical models introducing further sources of uncertainty.

3.2.3.3.2 PV SIMULATION

Many models for PV simulation have been developed and are mainly used in the context of planning and sizing of PV systems and for yield estimation. They generally meet the requirements for PV power prediction for most conventional PV technologies. Application of these models needs the knowledge of the specification of module and inverter characteristics, as well as the description of the orientation and tilt angle of the modules. Although these parameters are usually at hand for specific systems, it is not straight forward for regional power forecasts, because detailed system information is generally not available for all – especially small – PV systems.

As an example, a PV simulation model for DC reported in [65] is briefly described. A basic advantage of this model is the applicability to both classic crystalline silicon and thin-film technologies. This meets the requirements for PV power forecasting, in particular also for regional forecasts, where generally various different module types have to be considered. The proposed model estimates the efficiency of a PV generator operating at maximum power point

(MPP) conditions with a two-stage approach with solar irradiance and module temperature as input. First, the basic influence of the irradiance for a module temperature of 25 °C on the efficiency is described with a parametric model (blue line in Figure 11). Second, the performance at module operating temperatures other than 25 °C is modeled by a standard approach using a single temperature coefficient. The module temperature may be approximated from the ambient temperature and irradiance on the module plane taking also into account the mounting type of the system. A typical PV efficiency characteristic is shown in Figure 11.

Figure 12: Example of normalized MPP efficiency as a function of irradiance on the module plane. Blue:

MPP efficiency for Tmodule = 25°C. Red: MPP efficiency for Tambient = 15 °C. Green: AC efficiency for

Tambient = 15 °C. The curves are normalized to the efficiency at standard test conditions (STCs). STCs are −2 defined by Tmodule = 25 °C, Irradiance = 1000 Wm , and the standard spectrum of air mass AM1.5.

The DC power output of a PV generator then can be calculated using the forecasted irradiance and temperature as input. The model requires specification of the model parameters, the installation type, and the nominal power. For grid-connected PV systems, also the inverter characteristics have to be considered in addition to the module performance. Usually, this is done by describing the efficiency as a function of DC power input. Figure 11 also shows an exemplary curve for overall conversion efficiency from irradiance to AC power.

For PV technologies other than c-Si or some thin-film technologies other models are necessary. For example, concentrating PV systems equipped with triple-junction cells show a strong dependency on the spectral distribution of solar irradiance and no standard model is yet available.

3.2.3.3.3 UPSCALING TO REGIONAL POWER PREDICTION

Renewable power prediction for utility applications usually requires forecasts of the cumulative power production for the particular control areas. For both, wind and solar power forecasts, this is routinely done by up-scaling from a representative set of single sites. There is almost no loss in accuracy by the up-scaling approach, given that the representative set approximates the basic properties of the total data set, in particular with respect to the spatial distribution of the installed power. Most small-scale irradiance variabilities are subject to spatial averaging effects and do not have to be considered for a regional forecast. Large-scale variations, for example by approaching frontal systems, which impact regional prediction, are taken into account by using representative systems, which show the same behavior than surrounding systems.

The quality of the up-scaling procedure depends on the definition of the representative data set, which should show a similar response to the irradiance conditions as the full ensemble. A correct representation of the spatial distribution of the nominal power is most important in this respect. In addition, the distribution of PV system orientations has an influence on the ensemble power production and the subset should reflect the original orientations. Finally, the mix of module types has to be considered due to the different part-load behavior of different module types. For example, thin-film technologies generally show a better efficiency for low irradiance values.

3.2.3.3.4 STATISTICAL METHODS

Statistical methods may be applied to forecast PV power directly from NWP output or previous measurements or to improve local or regional PV power forecasts derived with explicit physical modeling. The methods are largely the same as for irradiance forecasts reported earlier ranging from simple regression techniques to more advanced AI techniques.

A fully statistical algorithm has been proposed by [18]. An auto-regressive model with exogenous input is used to predict PV power on the basis of measured power data and NWP global irradiance predictions. The weight for both input data sources is adjusted depending on the forecast horizon, ranging from one hour to two days. The model coefficients are obtained by recursive least-squares fitting with forgetting using online measured data in order to cope with temporal changes that affect PV power output, for example, snow covered modules or temporal shading effects. A statistical method for the combination of forecasts from different NWP models to derive optimized PV power predictions is presented in [66]. The method uses measured PV power and measured and predicted meteorological parameters of various NWP models as input to a statistical procedure based on the classification of different weather situations.

3.2.4 EVALUATION OF SOLAR IRRADIANCE AND SOLAR POWER FORECASTS

Analyzing and specifying the forecast accuracy is a major step towards a valuable solar power forecast. A good knowledge of forecast accuracy is the basis for any decision and assists in choosing between different forecasting products. In research, forecast evaluation is necessary for model testing and further model development.

To assess the forecast‘s accuracy, it is compared with the corresponding measured irradiance or solar power values. Many different aspects can contribute to forecast evaluation. A comprehensive overview of forecast verification methods is given in [57]. The most commonly used evaluation methods in solar power forecasting are summarized in the following.

The examples shown use a data set consisting of measurement data as well as a corresponding set of forecast data. Measured data are hourly global irradiance values from a station of the German Weather Service (DWD) in Mannheim, Germany (49.2° N, 9.56° E; station height: 96 m). The evaluations are done for the period from 1 January 2007 to 31 October 2007. Forecast data are based on the 0:00 UTC model run of the ECMWF deterministic global model with a spatial resolution of 25 km x 25 km and a temporal resolution of 3 h in combination with the post-processing procedure as described in [35].

3.2.4.1 EVALUATION TECHNIQUES

3.2.4.1.1 GRAPHICAL ANALYSIS

Graphical comparisons of forecasts with measured values give a first and intuitive impression of forecast accuracy and provide a detailed insight into the forecast performance. This is especially important from the scientific point of view for understanding the reasons for forecast errors and improving the forecast.

Time series of predicted irradiance in comparison with measured irradiance visualize forecast quality in a comprehensive way (Figure 12, left). The example illustrates the high forecast accuracy in clear-sky situations and larger errors in broken cloud situations. Scatter plots of predicted over measured irradiance are another graphical option to examine the forecast quality (Figure 12, right). These plots can reveal systematic deviations depending on the irradiance conditions and show the range of deviations that are related to the forecasts.

Figure 13: Left: Time series of predicted and measured global irradiance for the period from 29 April 2007 to 6 May 2007. Right: Scatter plot of predicted vs. measured global irradiance.

3.2.4.1.2 STATISTICAL ERROR MEASURES

A quantitative description of forecast accuracy can be obtained using statistical error measures. For the evaluation of wind and solar power forecasts, it is common practice to use the root mean square error (RMSE) as a main score [67][68]. As additional error measures the standard deviation of the error and the mean absolute error (MAE) are used. The MAE is appropriate for applications with linear cost functions, i.e., where the costs resulting from a poor forecast are proportional to the forecast error. The RMSE is more sensitive to large forecast errors, and hence is suitable for applications where small errors are more tolerable and larger errors cause disproportionately high costs, as, e.g., in case of utility applications.

For applications where decisions are related to threshold values, the additional investigation of the frequency distribution of the forecasts may be useful. The agreement of the distribution functions of measured and predicted time series can be evaluated, for example, using the Kolmogorov-Smirnov test integral [69].

Statistical error measures for solar irradiance or solar power differ significantly depending on whether they are based on daylight hours only or on all 24 hours of a day. Within the solar resource assessment community, usually only daytime values (with non-zero irradiance) are considered for accuracy assessments. However, parts of the electric utility sector are used to evaluate energy production forecasts including all hours of the day, resulting in lower RMSE values. Evaluation results shown in this report use daylight hours as a basis, unless noted otherwise.

Also, error measures for wind and solar power predictions are usually normalized and, e.g., the relative RMSE is used. As reference values both, the mean value of irradiance or solar power and the installed power of the plant are used. The latter is used frequently in the electric utility sector. Obviously the respective choice results in a significantly different error measure. The results shown in this report are based on the mean value as reference, unless indicated otherwise.

3.2.4.1.3 REFERENCE FORECASTS AND SKILL SCORES

In addition to statistical error measures the use of a – mostly trivial – reference forecast is frequently applied. The improvement relative to this reference then serves as another measure of forecast quality. The most common reference for short-term forecasts is persistence and different concepts for its use have been introduced.

A specific feature of solar power forecasting is the deterministic component of solar irradiance due to the geometrically determined path of the sun. This characteristic may be added as a constraint to the most simple form of persistence. The simplest way to achieve this is to consider the measured value of the previous day at the same time as a forecast value. In this case no additional information about the daily pattern is necessary. However, the daily variability of solar irradiance leads to a non-optimal approach when applied to forecast horizons of some hours only. A combination of persistence of the dimensionless clear-sky index (ratio of global irradiance to irradiance in a well- defined clear-sky situation) with a clear-sky model better accounts for the deterministic daily pattern of irradiance and results in smaller persistence errors. The same concept may be applied for forecast horizons of several days ahead using the average value of the clear-sky index for all daylight hours of the previous day.

If a climatological value of the clearness index is used instead of the observed one the same method can be used resulting in a climate average, which is obviously not depending on the forecast horizon.

A comparison of the ECMWF-based forecast with two reference models is given in Figure 13 (left). Persistence up to a few hours ahead is compared to the intraday ECMWF-based forecast. After 2 h, the ECMWF-based forecast performs significantly better than persistence. For the prediction of hourly irradiance values, for example, 3 h ahead, it is therefore preferable to use the NWP-based intraday irradiance forecast than extrapolating measured irradiance values. For forecasting the next hour, persistence is superior to the intraday NWP forecasts. For forecast horizons ahead of two days, climate averages show lower errors and should be preferred.

Figure 14: Relative RMSE of ECMWF-based forecasts and of reference models depending on the forecast horizon.

Commonly the performance of a forecast model compared to a simple reference model is described by the use of a skill score. Skill scores define the difference between the forecast and a reference forecast normalized by the difference between a perfect and the reference forecast:

Skill score = (Score - Score ref) / (Score perf - Score ref).

Its value thus ranges between 1 (perfect forecast) and 0 (reference forecast). A negative value indicates a performance even worse compared to the reference.

Skill scores may be applied not only for comparison with a simple reference model but also for intercomparisons of different forecasting approaches (improvement scores).

3.2.4.1.4 ADDITIONAL FACTORS AFFECTING FORECAST ACCURACY

Actual solar irradiance is mainly determined by geometric solar elevation and atmospheric cloud amount and distribution. An analysis of the forecast quality in dependence on these parameters therefore provides valuable additional information.

Variable cloud situations are generally more difficult to predict and therefore show larger forecast errors than clear-

Figure 15: Bias (left) and standard deviation of the error (right) of the irradiance forecast in Wm-2 as a function of the cosine of the solar zenith angle (cos ΘZ) and the clear-sky index kt*. sky situations. Also, the solar zenith angle ΘZ determines the maximum achievable irradiance and hence influences the magnitude of forecast errors. Results of an error assessment depending on these two parameters are shown in Figure 14, where the clear-sky index is used as an indicator of the cloud situation.

Solar irradiance is overestimated for intermediate cloud situations (0.3 < kt* < 0.8), which in its magnitude also depends on the solar zenith angle. For situations predicted as overcast (kt* < 0.2) the observed irradiance is underestimated, whereas for clear-sky situations no systematic deviations occur. These results may be used for a situation-dependent bias correction.

The standard deviation of the forecast error (Figure 14, right) shows only small values for situations predicted as clear sky (kt* ≈ 1) and for overcast situations with low irradiance (kt* < 0.2). The largest deviations between observation and forecast occur for situations with variable cloud cover, characterized by intermediate values of the clear-sky index. Again, the strong influence of the solar elevation is obvious. Knowledge of this dependency may be used for the calculation of situation-specific uncertainty information.

3.2.4.1.5 UNCERTAINTY INFORMATION

Specifying the expected uncertainty of solar irradiance or power predictions is always a valuable if not necessary addition to any forecast. As a basis for quantifying the expected uncertainty of a forecast, the probability distribution function of forecast errors or – equivalently – irradiance or power predictions have to be known. From this information, confidence or error intervals can be derived that indicate the range in which the actual value is expected to appear with a quantified probability.

Two different approaches are common to provide uncertainty information: • ensemble prediction systems • analysis of simultaneous historical time series of forecasts and observations.

In ensemble prediction systems, several runs of the same model are computed at the same time starting from slightly different initial conditions [70]. Due to the high complexity and nonlinearity of atmospheric processes, even small changes to the initial conditions can change the results significantly and each of these runs will result in a different forecast. The forecast output then is usually interpreted in terms of a probability density function, i.e., a direct measure of uncertainty.

Apart from ensemble predictions, uncertainty information may be derived from the analysis of historical time series of forecasted and observed data. In [35], weather-specific error intervals are determined on the basis of a detailed error analysis. If a normal distribution of forecast errors for different cloud and solar elevation situations and a negligible bias are assumed, the distribution function of errors is completely described by the standard deviation of the forecast errors. Applied to defined classes of clear-sky index and cosine of the solar zenith angle (see Figure 14, right), specific values of the standard deviation of the error are determined depending on these two parameters and described by an analytical function with good accuracy. Confidence intervals with a given uncertainty level then can be given. Figure 15 (left) shows forecasted irradiance values with 95% confidence intervals derived by this method in comparison with observed irradiance. In clear-sky situations the forecast quality is high and prediction intervals are narrow. On days with variable cloudiness, large deviations are to be expected for forecasts with hourly resolution

Figure 16: Forecast of global irradiance Iglob with confidence intervals of an uncertainty level of 95% compared with measured irradiance for six days in May 2007 for a single site (left) and for the average of two hundred measurement stations in Germany (right). for single stations. Regional forecasts show an improved agreement between forecasts and measurements, with narrow confidence intervals for different weather situations (Figure 15, right).

A further approach to determine the probability distribution of forecast values is presented in [18]. The authors propose the use of quantile regression to determine the probability distribution of normalized power values in dependence on predicted normalized power. The advantage of this method is that no assumption on distribution functions is necessary. However, the method is computationally more expansive because of the separate calculation of different uncertainty levels.

3.2.4.2 EVALUATION OF DIFFERENT APPROACHES TO IRRADIANCE FORECASTING

In the framework of the International Energy Agency‘s SHC Task 36 “Solar Resource Knowledge Management” [71] a common benchmarking procedure for solar irradiance forecasting has been developed and applied to compare irradiance forecasting procedures of seven participants of IEA SHC Task 36. The different methods include the use of meso-scale NWP models, the application of statistical post-processing tools to forecasts of NWP models, and also one approach involving the meteorologist’s expert knowledge to interpret and combine forecast data of different sources.

Up to now, there are no operational direct irradiance forecasts available from NWP (with exception of the ECMWF- IFS, but only since 2011). Therefore, there are no benchmarking results available with respect to the direct component or the DNI. This section concentrates on global irradiances instead.

A common measurement data set has been chosen for the benchmark consisting of hourly global irradiance values from four different climatic regions within Europe: Southern Germany, Switzerland including mountainous stations, Austria, and Southern Spain. The evaluation period was July 2007 until June 2008.

Hourly global irradiance forecasts up to 3 days ahead provided by participants of IEA SHC Task 36 are included in the benchmarking. The different forecasting approaches are all based on global scale NWP model predictions, either ECMWF-IFS global model forecast data or GFS data. For the local refinement of these coarse resolution forecasts, different methods are proposed within IEA SHC Task 36.

The algorithms used in the different forecasting methods can be grouped into four categories [49]: • Combination of a global scale NWP model with a post-processing technique involving historical surface observations or satellite-derived irradiance data (ECMWF/Oldenburg University, BlueSky, Meteomedia-MOS) • Combination of a meso-scale NWP model and a post-processing technique based on historical surface observations (CENER, Ciemat) • Forecasts of the meso-scale model WRF without any integration of observation data (Meteotest, Jaén University) • Human interpretation of NWP output (BlueSky-HI).

Figure 17: RMSE of the five forecasting approaches and persistence for all three German stations for the first three forecast days (left) and for the single stations for the first forecast day only (right). The average hourly global irradiance for all stations including only daylight hours was 227 Wm−2. Five of these forecasting methods have been evaluated with data from three German ground stations. The relative RMSE in dependence on forecast horizon and location is shown in Figure 16. With respect to RMSE the forecasting approaches combining a global model with post-processing (ECMWF/OL, BlueSky, Meteomedia-MOS) perform similarly and show consistently better results than mesoscale models. The bias is mostly small for all evaluated

forecasting approaches. All approaches show a significant improvement in comparison with persistence for all forecast horizons. The seasonal dependency of forecast errors (Figure 17) shows that during winter months with low solar elevations and low clear-sky irradiances the absolute errors are small and relative errors are large. Generally, the forecasts derived from global NWP models show a higher accuracy than those based on meso-scale models with a strong dependency of the difference on the month of evaluation. Furthermore, the accuracy of the meso-scale model forecasts shows a stronger dependence on the site of evaluation.

Figure 18: Absolute (left) and relative (right) global irradiance forecast errors: RMSE (solid line with circles) and bias (dashed lines) of five different forecasting approaches and persistence in dependence on the month for the first forecast day using data from three German sites.

Results for a South European site (Spain) are shown in Figure 18. The intercomparison shows similar trends: The approach using a global model in combination with post-processing (ECMWF/Oldenburg) is superior to the meso- scale models (CENER, Ciemat). The WRF forecast processed at University of Jaén without post-processing but using historical measurement data outperforms the other meso-scale models.

Figure 19: RMSE for the first, second, and third forecast day for stations in Spain. The average hourly global irradiance including only daylight hours was 391 Wm−2.

The evaluation shows that forecast accuracy strongly depends on the meteorological situation and clear-sky days generally show smaller forecast errors than cloudy days. Therefore, e.g., in Southern Spain smaller absolute errors

occur compared to Central Europe. Concerning relative errors, normalization to the higher average irradiance results in further reduced values in Spain. Detailed results of the forecast evaluations are given in [49].

3.2.4.3 EVALUATION OF SUB-DAILY IRRADIANCE FORECASTS

Shortest-term global irradiance forecasts up to 6 hours may be derived using satellite data [29] (see also chapter 2.3.2.2). An evaluation with comparisons with the ECMWF-OL forecasts and persistence has been performed using irradiance measurements from more than hundred German stations for a three months period (July – September 2011). Only hours in which satellite-based forecasts are available for all forecast horizons up to 5 h ahead are included in the evaluation restricting the evaluation to afternoon hours because necessary cloud information is derived from the visible channel of Meteosat-8 only.

The benefits of the different data sources and approaches for the various forecast lead times are shown in Figure 19. Up to several hours, the satellite-based approach is superior to the other methods. Compared with NWP-based methods, the improvement is larger for regional averages than for single sites. Ahead of 6 h, NWP-based forecasts are the best choice. Depending on the spatial average, persistence of ground-measured clear-sky index values performs better than the NWP-based forecasts up to 2 h (single sites) or 3 h (regional average) ahead. The results are in agreement with a similar comparison for the United States [30].

Figure 20: Relative RMSE of global irradiance forecast for single sites (left) and regional average values (right) based on motion vectors from satellite images (orange) compared with global model forecasts ECMWF-OL (dark blue), satellite-based irradiance values (light blue), and persistence (red).

Forecasts in the sub-hour range (not shown here) are likely to benefit from any combination of persistence, time series models, and information from ground-based sky imagers. A lower limit for the accuracy of satellite-based forecasts is given by the inherent uncertainty of the irradiance retrieval from satellite data (light blue line in Figure 19).

3.2.5 EXAMPLE OF A REGIONAL PV POWER PREDICTION SYSTEM

3.2.5.1 BACKGROUND & OVERVIEW

As one of the PV power prediction services available on the market [1][2][3][66] the joint operational system of Oldenburg University and Meteocontrol GmbH [4][36] is presented briefly. The PV power prediction scheme provides hourly forecasts up to three days ahead and combines explicit physical modeling with statistical tools at different stages of the modeling chain. Measurements from several thousand PV systems monitored by Meteocontrol GmbH in Germany are used for statistical post-processing, continuous evaluation, and further development.

Evaluation and post-processing of irradiance forecasts is based on hourly measurements of irradiance from more than 200 meteorological stations in Germany for the period January 2007 – October 2007. PV power forecasts are evaluated with hourly power output data for the period October 2009 – September 2010 within the control areas of the German transmission system operators ‘50 Hertz’ and ‘TenneT’.

For the evaluation of regional forecasts, the actual regional power production has to be estimated from available measurements by up-scaling from a representative data set to the overall ensemble. This requires a correct representation of the systems in particular with respect to the spatial distribution of the installed power. An evaluation of the up-scaling process is done using the regional power forecasts for the control area of 50Hertz, where 77 representative PV systems are compared to measurements of all (i.e., more than 500) monitored systems.

A mostly automatic quality control [4] was applied to the measured power data. Main components are time synchronization, filtering of data identified as wrong, and handling of zero power values, resulting e.g. from system failures or snow coverage, during daytime.

The PV power prediction system is based on irradiance and temperature forecasts of the ECMWF global model up to three days ahead. Figure 10 illustrates the different steps to derive PV power forecasts from ECMWF irradiance forecasts: Site-specific hourly forecasts are derived from the low-resolution ECWMF forecasts for the representative system sites and a post-processing procedure using measured irradiance data. Then the forecasts of global horizontal irradiance are converted to the module plane with a tilted irradiance model. A PV simulation model produces the PV power forecasts from the forecasts of tilted irradiance and further statistical post-processing is applied including additional meteorological parameters. Finally, the regional forecasts are obtained by up-scaling the power production from the representative set of PV systems.

3.2.5.2 SOLAR IRRADIANCE FORECASTS

The global irradiance forecasts from the ECMWF global model in the version used here have a temporal resolution of three hours and a spatial resolution of 25 km x 25 km. The implemented radiative transfer scheme is described in [37]. An optimized hourly and site-specific irradiance forecast is derived from the ECMWF output applying several steps [35]: First, a spatial averaging over a region of 100 km x 100 km is applied. This size has shown a maximum forecast accuracy in a comparison of different averaging areas. For the temporal interpolation – performed on the clear-sky index – the combination of the forecast data and a clear-sky model is used to account for the diurnal irradiance pattern. Finally, the weather-specific bias correction described earlier is applied to these forecasts. The correction function is continuously updated using measured irradiance values of the last 30 days. For evaluation, it was fitted on a training data set and evaluated on a test set.

An evaluation of the forecast accuracy depending on the forecast horizon resulted in the following figures: For single sites the relative RMSE amounts to 36.9% for the first forecast day and increases to 46.3% for the third forecast day. Spatial averaging effects lead to a significantly higher accuracy for an ensemble of distributed stations and the relative RMSE for the ensemble of all considered stations in Germany amounts to 13.4% and 22.5% for the first and third forecast day, respectively.

The accuracy of regional forecasts for arbitrary ensembles of stations depends on the size of the region to be considered. The reduction of errors when considering an ensemble of stations instead of a single station is determined by the cross-correlation of forecast errors of the systems in the ensemble. The correlation coefficient of

the forecast errors of two stations depends on the distance between the stations and is described by an exponential function [35] (Figure 20, left). This model in combination with a statistical approach to derive the expected errors of mean values is used to estimate forecast errors RMSEensemble for arbitrary scenarios of ensembles of stations. Figure 20 (right) shows the error reduction factor f = RMSEensemble/RMSEsingle for different ensembles of stations over the size of the region where the stations are distributed. For comparison, the error reduction factor determined directly from the data is given. For example, ensembles of sites equally distributed over a 3°x3° region show a RMSE of the

Figure 21: Left: Correlation coefficient of forecast errors of two stations depending on the distance between both. Blue dots: observations, red dots: exponential fit function. Right: Error reduction factor f = RMSEensemble/RMSEsingle for regions with increasing size. forecast reduced to about half of the RMSE for a single site.

3.2.5.3 PV POWER FORECASTS

To produce a regional PV power forecast the predicted horizontal irradiance has first to be converted to the PV module plane. Numerous models have been presented for this step. A combination of the diffuse fraction model of Skartveit and Olseth [62] and the diffuse sky model proposed by Klucher et al. [64] is widely applied, the latter accounting for the anisotropic effects of horizon brightening and circumsolar enhancement. The module irradiance then has to be transferred to electric power output via a PV simulation model [65] making also use of corresponding temperature forecasts. Further, models of the efficiency characteristics of the inverter and the different system losses are applied [4].

Further post-processing is necessary for a proper up-scaling of the single site forecasts from the representative subset to the region to be considered. The representativeness of the subsets for the two control areas used for evaluation is assessed by analyzing the basic properties of these subsets in comparison with the corresponding properties of the overall data sets [4]. Information on the spatial distribution of the overall installed power is based on the data published by the transmission system operators while representative information on the module type and geometric orientation of the PV systems is taken from an available monitoring database. To adjust for differences in the spatial distribution of the installed PV capacity between the representative subsets and the complete set of stations, a weighting procedure based on the ratio of the installed capacities of both the complete and representative sets is introduced.

In addition to these routine post-processing steps, certain weather-dependent situations call for specific post- processing corrections, for example, fog situations and snow-covered modules. Especially snow cover constitutes a severe source of error in PV power calculation due to the reduction of PV power to almost zero even in case of high solar irradiance. Given that the PV power forecast reflects the irradiance forecast this situation results in a strong overestimation of PV power. Empirically, the PV power prediction may be improved during periods of snow cover when measured PV power production data and additional meteorological parameters are included [36]. Here, forecasts of temperature and snow depth as well as the observed snow depth on the previous day have shown to be of systematic value to PV power forecasts.

3.2.5.4 EVALUATION OF LOCAL AND REGIONAL PV POWER FORECASTS

The basic accuracy measures for different forecast versions and persistence are summarized in Table 2 for intraday and day-ahead forecast horizons for the two control areas. All error values are normalized to the nominal power as this is common practice in large parts of the energy industry. The evaluation includes all 24 hours of the day.

The basic forecasting approach uses a simple up-scaling scheme and does not include a correction for snow covered modules. More sophisticated post-processing techniques (detailed up-scaling, snow correction) eliminate the bias and reduce the RMSE error between 0,5 % and 0,9% and show the potential of introducing post-processing techniques. The effect of increased forecast accuracy due to spatial smoothing is indicated by the smaller RMSE values for the larger control area of TenneT.

Forecast errors for single PV systems are significantly larger. The RMSE values are nearly twice as high compared to the values for the control areas (Table 3).

Table 2: Relative bias and relative RMSE of different regional forecasting approaches for the German control areas of 50 Hertz and TenneT (for all 24 h of the day).

rel. bias rel. RMSE

Intraday [%] Day-ahead [%] Intraday [%] Day-ahead [%] 50 Hertz

Basic 0,7 0,6 4,8 5,3

Detailed upscaling & snow correction 0,0 0,0 3,9 4,6

Persistence 0,0 0,0 8,2 10,2 TenneT

Basic 0,6 0,6 4,3 4,6

Detailed upscaling & snow correction 0,2 0,1 3,6 4,1

Persistence 0,0 0,0 6,8 8,8

Table 3: Relative bias and relative RMSE of different forecasting approaches for single PV systems within the control area of 50 Hertz (for all 24 h of the day).

rel. bias rel. RMSE

Intraday [%] Day-ahead [%] Intraday [%] Day-ahead [%]

Basic 0,5 0,6 8,1 8,6

Snow correction 0,2 0,1 8,0 8,5

Persistence 0,0 0,0 12,4 14,0

3.2.6 SUMMARY AND OUTLOOK

Solar power forecasting will be an essential component of any future energy supply system which uses large amounts of fluctuating solar power. Today, increasing effort is spent on forecasting of solar irradiance and solar electricity generation as a consequence of the need for precise and detailed forecast data in the energy sector.

Different forecasting approaches have been introduced up to now. Numerical weather prediction (NWP) based techniques with typical forecast horizons up to 2 days ahead represent the mainstream of solar power forecasting due to its importance for most grid integration issues. For shorter time scales from 30 minutes to 6 hours, forecasts based on cloud motion vectors from satellite images have shown a promising performance. Below one hour ground-based observations from e.g. sky imagers may provide valuable information whereas for the very short term time scale from minutes up to a few hours’ time series models using on-site irradiance measurements have been introduced. Independent of the chosen forecast technique a detailed evaluation is a key to model testing and further model development. A proper accuracy assessment provides valuable information for all users that rely on the forecasts as a basis for decision making.

Improvements in NWP-based irradiance forecasting may be expected as these models will develop with respect to resolution, parameterizations, and data assimilation. In any case, modeling and parameterization of clouds and radiation in both global and meso-scale NWP models as well as the integration of enhanced aerosol information play a key role in future improvements. Additionally, the extension of operational NWP forecasts towards direct or direct normal irradiance is highly recommended. A further promising approach is the use of ensemble prediction systems. Apart from the inherent uncertainty information, ensemble prediction systems perform superior to deterministic models for forecast horizons longer than typically 1 or 2 days ahead.

Since all modeling approaches finally will show remaining errors, both systematic and stochastic, statistical post- processing techniques will continue to contribute to future enhancement in solar power prediction. The availability of high-quality and up-to-date measurement data of solar irradiance and solar power will be of critical importance. Statistical modeling may also be applied for an optimal combination of different data sources, e.g., on-site measurements, satellite-based forecasts, and output of one or several NWP models. Depending on the time scale of interest, different combinations of these data sources may be chosen.

References to § 3.2:

[1] Bofinger S and Heilscher G (2006) Solar electricity forecast: Approaches and first results. In: Proc. 21st European Photovoltaic Solar Energy Conf., 2641–2645. Dresden, Germany. [2] Remund J, Schilter C, Dierer S, et al. (2008) Operational forecast of PV production. Proc. 23rd European Photovoltaic Solar Energy Conf., Valencia, Spain, 3138–3140.

[3] Le Pivert X, Sicot L, and Merten J (2009) A tool for the 24 hours forecast of photovoltaic production. Proc. 24th European Photovoltaic Solar Energy Conf., Hamburg, Germany, 4076–4079. [4] Lorenz E, Scheidsteger T, Hurka J, et al. (2010) Regional PV power prediction for improved grid integration. Progress in Photovoltaics: Research and Applications. [5] Perez R, Kmiecik M, Schlemmer J, et al. (2007) Evaluation of PV generation capacity credit forecast on day- ahead utility markets. In: Proceedings of the ASES Annual Conference. Cleveland, OH, USA. [6] Wittmann M, Breitkreuz H, Schroedter-Homscheidt M, and Eck M (2008) Case studies on the use of solar irradiance forecast for optimized operation strategies of solar thermal power plants. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 1, 18–27. [7] Takayama S, Iwasaka Y, Hara R, et al. (2009) Study on scheduling of PV power station output based on the solar radiation forecast. Proc. 24th European Photovoltaic Solar Energy Conf., Hamburg, Germany, 4127– 4131. [8] Kudo M, Takeuchi A, Nozaki Y, et al. (2009) Forecasting electric power generation in a photovoltaic power system for an energy network. Electrical Engineering in Japan 167, 16–23. [9] Rikos E, Tselepis S, Hoyer-Klick C, and Schroedter-Homscheidt M (2008) Stability and power quality issues in microgrids under weather disturbances. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 1, 170–179. [10] Yamamoto S, Park JS, Takata M, et al. (2003) Basic study on the prediction of solar irradiation and its application to photovoltaic–diesel hybrid generation system. Solar Energy Materials & Solar Cells, 75, 577– 584. [11] Traunmüller W and Steinmaurer G (2010) Solar irradiance forecasting, benchmarking of different techniques and applications to energy meteorology. Proc. EuroSun 2010. Graz, Austria. [12] Ernst B, Oakleaf B, Ahlstrom ML, et al. (2007) Predicting the wind. IEEE Power & Energy Magazine 5, 78–89. [13] Breitkreuz H, Schroedter-Homscheidt M, Holzer-Popp T, and Dech S (2009) Short range direct and diffuse irradiance forecasts for solar energy applications based on aerosol chemical transport and numerical weather modeling. Journal of Applied Meteorology and Climatology 48, 1766–1779. [14] Breitkreuz, H, Schroedter-Homscheidt, M, and Holzer-Popp, T (2007) A case study to prepare for the utilization of aerosol forecasts in solar energy industries. Solar Energy 11, 1377-1385. [15] Mayer B and Kylling A (2005) Technical note: The libRadtran software package for radiative transfer calculations – Description and examples of use. Atmospheric Chemistry and Physics, 5, 1855–1877. [16] Chowdhury BH (1990) Short-term prediction of solar irradiance using time-series analysis. Energy Sources 12, 199–219. [17] Reikard G (2009) Predicting solar radiation at high resolutions: A comparison of time series forecasts. Solar Energy 83, 342–349. [18] Bacher P, Madsen H, and Nielsen HA (2009) Online short-term solar power forecasting. Solar Energy 83, 1772–1783. [19] Mellit A (2008) Artificial intelligence technique for modelling and forecasting of solar radiation data: A review. International Journal of Artificial Intelligence and Soft Computing, 1, 52–76. [20] Kemmoku Y, Orita S, Nakagawa S, and Sakakibara T (1999) Daily insolation forecasting using a multi-stage neural network. Solar Energy, 66, 193–199. [21] Sfetsos A and Coonick AH (2000) Univariate and multivariate forecasting of hourly solar radiation with artificial intelligence techniques. Solar Energy, 68, 169–178. [22] Cao JC and Cao SH (2006) Study of forecasting solar irradiance using neural networks with preprocessing sample data by wavelet analysis. Energy 31, 3435–3445. [23] Chaabene M and Ben Ammar MN (2008) Euro-fuzzy dynamic model with Kalman filter to forecast irradiance and temperature for solar energy systems. Renewable Energy 33, 1435–1443. [24] Hocaoglu FO, Gerek ON, and Kurban M (2008) Hourly solar radiation forecasting using optimal coefficient 2- D linear filters and feed-forward neural networks. Solar Energy 82, 714–726.

[25] Voyant C, Muselli M, Paoli C, et al. (2009) Predictability of PV power grid performance on insular sites without weather stations: Use of artificial neural networks. Proc. 24th European Photovoltaic Solar Energy Conf., Hamburg, Germany. 4141–4144 [26] Cao JC and Lin XC (2008) Study of hourly and daily solar irradiation forecast using diagonal recurrent wavelet neural networks. Energy and Conversion Management 49, 1396–1406. [27] Beyer HG, Costanzo C, Heinemann D, and Reise C (1994) Short range forecast of PV energy production using satellite image analysis. Proc. 12th European Photovoltaic Solar Energy Conf., 1718–1721. Amsterdam, The Netherlands. [28] Hammer A, Heinemann D, Lorenz E, and Lückehe B (1999) Short-term forecasting of solar radiation: A statistical approach using satellite data. Solar Energy, 67, 139–150. [29] Lorenz E, Heinemann D, and Hammer A (2004) Short-term forecasting of solar radiation based on satellite data. Proc. EuroSun 2004, 841–848. Freiburg, Germany. [30] Perez R, Kivalov S, Schlemmer J, et al. (2009) Validation of short and medium term operational solar radiation forecasts in the US. Proc. ASES Annual Conference. Buffalo, NY, USA. [31] Taniguchi H, Otani K, and Kurokawa K (2001) Hourly forecast of global irradiation using GMS satellite images. Solar Energy Materials & Solar Cells, 67, 551–557. [32] Menzel WP (2001) Cloud tracking with satellite imagery: From the pioneering work of Ted Fujita to the present. Bulletin of the American Meteorological Society 82, 33–47. [33] Hammer A, Heinemann D, Hoyer C, et al. (2003) Solar energy assessment using remote sensing technologies. Remote Sensing of Environment 86, 423–432. [34] Chow CW, Urquhart B, Lave M, et al. (2011) Intra-hour forecasting with a total sky imager at the UC3 San Diego solar energy testbed. Solar Energy. [35] Lorenz E, Hurka J, Heinemann D, and Beyer HG (2009) Irradiance forecasting for the power prediction of grid- connected photovoltaic systems. IEEE Journal of Special Topics in Earth Observations and Remote Sensing 2, 2–10. [36] Lorenz E, Heinemann D, and Kurz C (2011) Local and regional photovoltaic power prediction for large scale grid integration: Assessment of a new algorithm for snow detection. Progress in Photovoltaics: Research and Applications. [37] Morcrette JJ, Barker HW, Cole JNS, et al. (2008) Impact of a new radiation package, McRad, in the ECMWF integrated forecasting system. Monthly Weather Review 136, 4773–4798. [38] Grell GA, Dudhia J, and Stauffer DR (1995) A description of the fifth-generation Penn State/NCAR mesoscale model (MM5). Technical Note NCAR/TN-398+STR, 121pp. Boulder, CO: National Center for Atmospheric Research. [39] Skamarock WC, Klemp JB, Dudhia J, et al. (2008) A description of the advanced research WRF version 3. Technical Note NCAR/TN-475+STR. Boulder, CO: Mesoscale and Microscale Meteorology Division, National Center for Atmospheric Research. [40] Heinemann D, Lorenz E, and Girodo M (2006) Forecasting of solar radiation. In: Dunlop ED, Wald L, and Suri M (eds.) Solar Resource Management for Electricity Generation from Local Level to Global Scale, 83–94. New York: Nova Science Publishers. [41] [Zamora RJ, Solomon S, Dutton EG, et al. (2003) Comparing MM5 radiative fluxes with observations gathered during the 1995 and 1999 Nashville southern oxidants studies. Journal of Geophysical Research 108(D2): 4050. [42] Zamora RJ, Dutton EG, Trainer M, et al. (2005) The accuracy of solar irradiance calculations used in mesoscale numerical weather prediction. Monthly Weather Review 133, 783–792. [43] Guichard F, Parsons DB, Dudhia J, and Bresch J (2003) Evaluating mesoscale model predictions of clouds and radiation with SGP ARM data over a seasonal timescale. Monthly Weather Review 131, 926–944. [44] Ruiz-Arias JA, Pozo-Vázquez D, Sánchez-Sánchez N, et al. (2009) Evaluation of two MM5-PBL parameterizations for solar radiation and temperature estimation in the south-eastern area of the Iberian Peninsula. Il Nuovo Cimento 31, 5–6.

[45] Sánchez-Sánchez N, Pozo-Vázquez AD, Ruiz-Arias JA, et al. (2007) An evaluation study of the MM5 solar radiation estimates in a complex topography area in southeastern Spain. Proc. 7th Annual Meeting of the European Meteorological Society. Madrid, Spain: San Lorenzo de El Escorial. [46] Remund J, Perez R, and Lorenz E (2008) Comparison of solar radiation forecasts for the USA. Proc. 23rd European Photovoltaic Solar Energy Conf., Valencia, Spain, 3141–3143. [47] Perez R, Beauharnois M, Hemker K, et al. (2011) Evaluation of numerical weather prediction solar irradiance forecasts in the US. Proc. ASES Annual Conference. Raleigh, NC, USA. [48] Lara-Fanego V, Ruiz-Arias JA, Pozo-Vázquez D, et al. (2011) Evaluation of the WRF model solar irradiance forecasts in Andalusia (southern Spain). Solar Energy. [49] Lorenz E, Remund J, Müller SC, et al. (2009) Benchmarking of different approaches to forecast solar irradiance. Proc. 24th European Photovoltaic Solar Energy Conf., Hamburg, Germany, 4199–4208. [50] Glahn HR and Lowry DA (1972) The use of model output statistics (MOS) in objective weather forecasting. Journal of Applied Meteorology 11, 1203–1211. [51] Jensenius JS and Cotton GF (1981) The development and testing of automated solar energy forecasts based on the model output statistics (MOS) technique. Proc. 1st Workshop on Terrestrial Solar Resource Forecasting and on the Use of Satellites for Terrestrial Solar Resource Assessment. Newark, NJ: American Solar Energy Society. [52] Guarnieri RA, Pereira EB, and Chou SC (2006) Solar radiation forecast using artificial neural networks in South Brazil. In: Proceedings of the 8th ICSHMO, pp. 1777–1785. Foz do Iguaçu, Brazil. [53] Kratzenberg MG, Colle S, and Beyer HG (2008) Solar radiation prediction based on the combination of a numerical weather prediction model and a time series prediction model. In: Proceedings of EuroSun 2008. Lisbon, Portugal. [54] Perez R, Moore K, Wilcox S, et al. (2007) Forecasting solar radiation: Preliminary evaluation of an approach based upon the national forecast database. Solar Energy, 81, 809–812. [55] Pelland S, Gallanis G, and Kallos G (in press) Solar and photovoltaic forecasting through post-processing of the global environmental multiscale numerical weather prediction model. Progress in Photovoltaics: Research and Applications. [56] Mathiesen P and Kleissl J (2011) Evaluation of numerical weather prediction for intra-day solar forecasting in the continental United States. Solar Energy, 85, 967–977. [57] Jolliffe IT and Stephenson DB (2003) Forecast Verification: A Practitioner’s Guide in Atmospheric Science. Wiley. [58] Ruiz-Arias JA, Pozo-Vázquez D, Lara-Fanego V, et al. A high-resolution topographic correction method for clear-sky solar irradiance derived with a numerical weather prediction model. Journal of Applied Meteorology and Climatology 2011, e-View. [59] Ruiz-Arias, JA, Alsamamra, H, Tovar-Pescador, J, and Pozo-Vázquez, D (2010) Proposal of regressive model for hourly diffuse solar radiation under all sky conditions. Energy Conversion Management, Vol. 51 (5), pp. 881 - 893. [60] Morcrette J-J, Boucher O, Jones L, Salmond D, Bechtold P, Beljaars A, Benedetti A, Bonet A, Kaiser, JW, Razinger M, Schulz M, Serrar S, Simmons AJ, Sofiev M, Suttie M, Tompkins AM, and Untch A (2009) Aerosol analysis and forecast in the ECMWF Integrated Forecast System. Part I: Forward modelling. J. Geophys. Res. 114, D06206, doi:10.1029/2008JD011235. [61] Suzuki H, Watanabe Y, and Wakao S (2008) Short-term PV output forecast using just-in-time modeling. Proc. 23rd European Photovoltaic Solar Energy Conf., 3406–3408. Valencia, Spain. [62] Skartveit A, Olseth JA, and Tuft ME (1998) An hourly diffuse fraction model with correction for variability and surface albedo. Solar Energy, 63, 173–183. [63] Perez R, Seals R, Ineichen P, et al. (1987) A new simplified version of the Perez diffuse irradiance model for tilted surfaces. Solar Energy, 39, 221–231. [64] Klucher TM (1979) Evaluation of models to predict insolation on tilted surfaces. Solar Energy, 23, 111–114.

[65] Beyer HG, Betcke J, Drews A, et al. (2004) Identification of a general model for the MPP performance of PV modules for the application in a procedure for the performance check of grid connected systems. Proc. 19th European Photovoltaic Solar Energy Conf., 3073–3076. Paris, France. [66] Schmelter J and Focken U (2011) Operationelle Erfahrungen mit kombinierten Solarleistungsvorhersagen für deutsche ÜNBs und VNBs. Proc. 26th Symposium Photovoltaische Solarenergie, 376–381. Bad Staffelstein, Germany. [67] Lange M and Focken U (2005) Physical Approach to Short-Term Wind Power Prediction. Berlin, Heidelberg, New York: Springer. [68] Nielsen TS, Madsen H, Nielsen HA, et al. (2006) Advanced statistical modelling and uncertainty assessment for wind power forecasting. Proc. European Wind Energy Conference. Athens, Greece. [69] Espinar B, Ramírez L, Drews A, et al. (2009) Analysis of different error parameters applied to solar radiation data from satellite and German radiometric stations. Solar Energy, 83. [70] Molteni F, Buizza R, Palmer TN, and Petroliagis T (1996) The ECMWF ensemble prediction system: Methodology and validation. Quarterly Journal of the Royal Meteorological Society, 122, 73–119. [71] http://www.iea-shc.org/task36

3.3 GRID MANAGEMENT11

3.3.1 INTRODUCTION

The management of a grid is a difficult task, even when only the load is variable and all power plant output can be scheduled. In this case, the future scheduled power just needs to exceed the predicted load (plus the load forecast error, which is usually small) plus a contingency for the largest block of power to suddenly drop. This way to run a model is called the (n-1) criterion, because the safety margin using n power plants is calculated such that the largest 1 power plant or transmission line can suddenly drop off the system without harming the delivery of electricity to the consumer. However, also ENTSO-E (the European Network of Transmission System Operators for Electricity) agrees that the (n-1) criterion in the new context might no longer be valid and needs to be replaced by a probabilistic view of the power system. In their R&D plan, they write about topic 9 “Innovative Tools and Approaches for the Pan-European Network Reliability Assessment”: “Today, the most important principle of transmission network planning and operation is to guarantee (n-1)-preventive security as a strict constraint. According to the increasing uncertainties linked to the variable renewable energy generation sources, the pan-European electricity market and the communities’ opposition to new transmission infrastructures, meeting this constraint is getting more and more difficult. There is a need to re-evaluate the constraint of (n-1)-security: Can this (n-1)-criterion be improved by other regulatory constraints for reliability performances?” [1].

In addition to the challenge of variability of the renewable resource comes the challenge of monitoring potential of the current state of the grid. When the system goes from a few large power blocks to very many distributed generators, like in the case of Denmark (see §4.2.7) , not all power generation is monitored equally well, and the likelihood of data feed outages increases with the number of power stations and with the number of actors on the power market. The number of actors might also have an impact on the forecast quality. In Spain [2], all wind farm owners have to provide their own prediction – but the sum of those predictions is usually worse than the prediction of the central Sipreólico tool run by Red Electrica de España REE, the Spanish TSO [3]. The differences can easily be over 600 MW, so the question for the TSO is what to believe? Recently, with the large number of installations, the maximum wind power gradient in the grid reached 1 GW/h – therefore, REE starts to use 15-min updates of the forecasts. Generally, a centrally run and maintained system will have a more homogeneous (and usually better) performance than forecasts which have to be bought from third parties in a market, and then reported to the system. Another advantage of a central system is that it can take spatio-temporal correlations of the error into account, and thereby correct the timing error in the NWP to some extent. See also the chapter on spatio-temporal correlations in Giebel et al., 2011 [4].

The operation of the power grid entails not only the balance between generation and demand; it also entails the transport of the energy from the generation to the consumer. While the direction of this power transport in the last decades was quite clear (e.g. from nuclear power stations in the countryside to the load centres in the cities), the variable nature of the renewable resource makes matters more complicated. Large-scale offshore wind in the North Sea, solar power in Germany and further south and other concentrated Renewable Energy Sources RES fluctuate driven by the weather, so that the transport of electricity is getting more complicated. The lines also are not built to include the new amounts of power, so some tools like DLRs are used.

TSOs have to provide available power reserves to keep the power system stability in a case of sudden loss of power. Power systems with high RES power penetration might have to take care about the correction of secondary and tertiary power reserves due to the variable nature of RES power, although in a geographically large system with a

11 Lead author: Gregor Giebel

spread out wind power feed, this might be not the case [3]. The added investment costs due to backup generation could be avoided when the power system is well interconnected with the neighbouring systems and all power imbalances can be corrected through power export/import. Investment costs in new interconnection lines are rather lower comparing the investments in new conventional plants. So, the enhancement of power system interconnection stands as one of the tools for the correction of power production forecast error from RES. However, permission for construction of new transmission line is getting more complicated recently due to increased awareness of environmental impact of transmission lines.

While the scheduling of the power system still is a task of the TSO, the financial aspects are mainly connected to the markets. In most European countries, the main market for power is a day-ahead market, having a market closure at noon to trade for the next 24h-day. So the financially most important forecasts are for 13-37 hours ahead, to come at 1100 hours each day.

References to §3.3.1:

[1] ENTSO-E: Research and Development Plan. European Grid towards 2020 Challenges and Beyond. First Edition, 23 March 2010 [2] Morales, G.G.: Forecasting for reserves and grid constraints. Workshop on Best Practice in the Use of Short- term Forecasting of Wind Power, Delft (NL), 25 October 2006 [3] Gerardo Gonzalez Morales, Red Electrica de Espana (Spanish TSO): Wind power prediction in the Spanish system operation. Talk on the 2nd Workshop on Best Practice in the Use of Short-term Forecasting, Madrid (ES), 28 May 2008 [4] Giebel G., Brownsword R., Kariniotakis G., Denhard M., Draxl C. The State-Of-The-Art in Short-Term Prediction of Wind Power A Literature Overview, 2nd Edition. Project report for the Anemos.plus and SafeWind projects. 110 pp. Risø, Roskilde, Denmark, 2011

3.3.2 ELECTRICITY TRANSPORT

The additional production variability due to the increasing number of RES brings new challenges to TSOs which are responsible for the power system stable operation from the technical point of view. TSOs are also obliged to provide negotiated energy transport through interconnecting lines and the supply to end users e.g. by following market operator agreements from the economical point of view.

A transmission system is affected by weather conditions from the physical point of view. It is also affected by the impact of the weather conditions on the users at the production and consumption levels. Some maintenance works, which require few weeks of outage, are strongly dependant on weather conditions: a weather forecast failure can delay maintenance works thus influencing the power system reliability.

The windy regions are usually located far from the urban areas with concentrated consumption. A system transmitting large amounts of power over long distances is very sensitive to stability: atypical example is the Swedish power system with concentrated power generation in the north of Sweden and concentrated consumption in the south. In such power system topology, small production/consumption imbalance can cause further generation loss. Consequently, production/consumption forecast imposes itself as a very important tool for TSOs for the optimization of transmission of electricity.

The DC transmission concept is an emerging technical solution for the stability and efficiency problems for long distances [16]. A power system on one side of the DC line is not affected by problems on the other end of the line, which is a big advantage for RES. Using DC transmission, the forecast error doesn't influence the power system

stability to such an extent compared to AC transmission: a larger error of the power prediction may be accommodated without consequences for the transmission power stability.

A transmission system has the main task to transmit power from the producers to consumers. However, other demands are imposed to TSOs: deployment of RES connected to distribution systems transforms them from passive consumers to active consumers/producers.

In order to fulfil all these requirements, various different measurements and data from all power system users (production units and consumers) are necessary. Furthermore, to obtain these data good communication links and software tools are necessary to process them and give a correct picture of the power system state. The transmission grid is then not any more rigid but a flexible and manageable “smart grid". "Smart grid" is supposed to be well "informed" about every power system user conditions and requirements in order to be able to complete all tasks. In such smart grid, a power forecast will be one of the key input data for successful power system operation.

Taking into consideration the enormous difficulties facing all TSOs for the construction of new transmission lines, it is rather surprising to realize that little has been done to improve the existing network use and efficiency for the benefit of a smooth grid integration of distributed renewable energy sources. It also decreases the need to construct new transmission lines. In the future, environmental and social requirements for new electrical lines comprise:

• online monitoring of transmission lines (temperature, wind, loads, etc.);

• introduction of new network components (e.g. phase shift transformers);

• use of Flexible AC Transmission Systems (FACTS) devices;

• upgrade of degraded components such as cables, lines, protections and transformers. All of these urgent requirements are to be implemented by the transmission grid operators [17]. An example is the DESERTEC concept of a "super DLR transmission system" for electricity transport from Northern Africa to Europe and Middle East. DESERTEC is proposed by the DESERTEC Foundation to make use of solar and wind energy: it aims at promoting the generation of electricity in Northern Africa, the Middle East and Europe (EU- MENA) using solar power plants, wind parks and the transmission of this electricity to the consumption centers. The realization of the DESERTEC concept in this region is pursued by the industrial initiative Dii. Under the DESERTEC proposal, concentrating solar power systems, photovoltaic systems and wind parks would be spread over the desert regions in Northern Africa like the Sahara desert. Produced electricity would be transmitted to European and African countries by a super grid of high-voltage direct current cables. It would provide a considerable part of the electricity demand of the MENA countries and furthermore provide continental Europe with 15% of its electricity needs [18]. The construction of the DESERTEC first 500 MW solar farm in Morocco is scheduled to start in 2012.

3.3.3 STORAGE

One of the tools to prevent power system instability induced by production variations and/or forecast errors from RES is the installation of storage facilities. The essence of storage is to shift energy through time, by stocking-up the surplus generated power and releasing it when electricity consumption exceeds power production.

It is expected to have increased storage facilities within power systems in the future when approaching smart grid concept. Storage facilities have two advantages: power system stability support and power savings. Taking into consideration that storage facilities "correct" power production variations and power forecast error, it could be defined that power production forecast and storage facilities are respectively “software” and "hardware" TSO tools for a power system stable operation.

Energy storage enables to optimize the operation strategy of the power system and allows to:

• Minimize deviations to participate in structured markets; • Contribute to the secondary and tertiary power reserves;

• Increase of RES contribution for the regulation capacity.

When hydro pumping storage is available, the appropriate methodologies able to identify the best combined RES/hydro pumping storage strategies should be used. In the absence of hydro energy resources, other storage techniques (see Energy Storage Association (http://www.electricitystorage.org) for a discussion of available storage technologies, their advantages and disadvantages) may also be helpful and should be investigated (e.g. H2/Fuel Cells, compressed air/gas, flywheels, etc.) [17].

3.3.3.1 STORAGE AND ALTERNATIVE PROBLEM FORMULATIONS

Energy storage technologies can be operated as standalone units within an interconnected power system (network energy storage) or in coordination with a renewable energy (RE) source to counterbalance variations in power generation (dedicated storage). The way storage is used gives rise to a range of application areas12:

Network energy storage

On a network-scale level, energy storage presents a promising method to increase the reliability and lower the operational cost of energy production. It allows the network operator to deal with short-term changes in electricity load by taking full advantage of renewable sources, without the need to dispatch more expensive carbon-based generators. A considerable number of studies investigate optimal design and operation strategies for networks that combine renewable generation and storage units ([1]; [3]; [4]; [7]; [14]; [6]; [2]). These studies analyze the fact that storage technologies can improve the operational and economic profile of networks with appreciable contribution by non-dispatchable generators, i.e. generators which cannot be scheduled but follow their own schedule due to the availability of solar or wind resources. Most important, they open the way to the wide-scale adoption of microgrids, with the ability to autonomously serve regional loads using multiple energy sources.

Another potential for storage, which has not received equal attention but is similar in spirit to the applications studied above, is its use in energy arbitrage trading. This amounts to making profit by adopting the following round-trip trading rule: buy cheap power in off-peak periods, store it and sell it back to the market when electricity prices return to their peak (see e.g. [5]).

Dedicated storage

The opportunities created by dedicated storage are extremely relevant for independent producers that participate in an electricity market. Energy producers relying on intermittent sources, such as wind and sun, are frequently faced with deviations between their submitted bids and the corresponding delivered power. If the production at the time of delivery is different than promised , the producer may incur considerable losses due to balancing operations depending on the overall power supply in the market. From a trader’s point of view, storage facilities act as a buffer to smooth unstable generation and minimize imbalance costs. To achieve this goal, one needs to dynamically adjust the energy storage size depending on the time when the power is delivered to the grid (dynamic sizing). A significant amount of storage-related studies appearing in the literature deals with the optimal control of imbalance costs (see e.g. [11]; [12]; [8]; [9]; [15]).

12 See [10] Chapter 4 and [3].

3.3.3.2 STORAGE AND POWER GENERATION FORECASTING

A common element in all studies exploring storage facilities is the adoption of some forecasting technique to predict the output capacity of flexible generators.

Power forecasting is considered as an essential ingredient to the optimal operation of a storage device under intermittency in the delivery of power. Inaccurate forecasts can result in poor estimations of the necessary storage levels and high unbalancing costs. Some authors also recognize the importance of having an accurate representation of the uncertainty affecting the predicted production at various timescales. This task is accomplished by forecasting models that assign a probability distribution over the whole range of production level ([9]). For dynamic storing problems, where energy allocation decisions are taken sequentially, errors in the forecasted output propagate throughout the scheduling period. Therefore, any autocorrelation in the observed error structure should also receive considerable attention by the decision-maker [15].

3.3.3.3 CHALLENGES FOR PRODUCTION FORECASTING MODELS

Despite the wide-scale use of power forecasting techniques, little has been yet said about the relative advantages of state-of-the-art forecasting models in storage-related applications. This particularly applies to the benefits from incorporating weather predictions and onsite measurements versus using a purely statistical approach. A plausible explanation for the documented literature “gap” could be given on the premises that - in this application domain - power prediction is only an intermediate step towards deriving optimal operational, scheduling and trading strategies. Besides, the accuracy of the forecasting model may not be so important, as storage devices can compensate for observed deviations between the forecasted and the actual production output (see [12]).

The final choice of the model depends on the objective function and the overall formulation of the optimization problem. In a simple version of the optimal sizing problem, the power plant operator may be concerned about increasing his average revenue. However, he may also schedule storage capacity with an emphasis on avoiding extreme losses, which may result from unexpected weather phenomena or too optimistic forecasts about future power levels. So, depending on how the problem is formulated, the operator may be interested in the behavior of the interior of the predicted output density function or its extreme tails (see e.g. [9] for an example of optimal storage scheduling taking into account the variance of the predicted wind power).

3.3.3.4 OPEN RESEARCH QUESTIONS

1) Investigating the performance of different weather prediction models in storage-related applications (optimal operation, dynamic sizing, value of storage, etc.). The main issue that should be addressed here is whether prediction models – in conjunction with storage opportunities - manage to reduce operating costs and increase the ability of the system to deliver the targeted output.

2) Quantifying the operational and economic benefits of storage facilities. Energy storing is not by default the most advantageous method of balancing the intermittency risk associated with renewables. In any case, it should be compared with alternative strategies and technologies developed for this purpose. For instance, on a grid-scale level, one could obtain a smoother aggregate output profile by combining multiple generation units with low-correlated production patterns (see also [19] and section 3.4, for a discussion). Apart from the presented opportunities for improving the overall system’s reliability, the benefits of storage facilities should be evaluated based on an array of additional factors, such as the installation/operational cost, the profitability of the energy storage cycle and the technological risk associated with adopting innovative and less-established storage devices."

References to §3.3.3:

[1] Bagen and R. Billinton (2005), “Evaluation of different operation strategies in small stand-alone power systems,” IEEE Trans. Energy Conversion 20, pp. 654–660. [2] Balamurugan, P. , S. Kumaravel and S. Ashok (2011), “Optimal Operation of Biomass Gasifier Based Hybrid Energy System”, ISRN Renewable Energy. [3] Barton, J. P. and D. G. Infield (2004), “Energy Storage and Its Use with Intermittent Renewable Energy”, IEEE Transactions on Energy Conversion 19 (2), pp. 441-448. [4] Barton, J. P. and D. G. Infield (2006), “A probabilistic method for calculating the usefulness of a store with finite energy capacity for smoothing electricity generation from wind and solar power”, Journal of Power Sources 162, pp. 943–948. [5] Bathurst, G. N. and G. Strbac (2003), “Value of combining energy storage and wind in short-term energy and balancing markets,” Electric Power Systems Research, 67(1), pp. 1–8. [6] Brekken, T.K.A., Yokochi, A., von Jouanne, A., Yen, Z.Z., Hapke, H.M., Halamay, D.A. (2011), "Optimal Energy Storage Sizing and Control for Wind Power Applications," IEEE Transactions on Sustainable Energy 2(1), pp.69-77. [7] Brown, P.D., J.A.P. Lopes and M. A. Matos (2008), “Optimization of Pumped Storage Capacity in an Isolated Power System with Large Renewable Penetration”, IEEE Transactions on Power Systems 23 (2), pp. 523 – 531. [8] Costa, L.M., Bourry, F., Juban, J. and Kariniotakis, G. (2008a), "Management of Energy Storage Coordinated with Wind Power under Electricity Market Conditions," in Proceedings of the 10th International Conference on Probabilistic Methods Applied to Power Systems (PMAPS '08). [9] Costa, L.M., Bourry, F., Juban, J. and Kariniotakis, G. (2008b), “A Spot-Risk-Based Approach for Addressing Problems of Decision-Making under Uncertainty", in Proceedings of the 10th International Conference on Probabilistic Methods Applied to Power Systems (PMAPS '08). [10] EWEA (2010), “Powering Europe: wind energy and the electricity grid”, the European Wind Energy Association, Technical Report. [11] Korpaas, M., A. T. Holen, and R. Hildrum (2003), “Operation and sizing of energy storage for wind power plants in a market system,” International Journal of Electrical Power and Energy Systems 25 (8), pp. 599–606. [12] Koeppel, G. and M. Korpaas (2006), “Using storage devices for compensating uncertainties caused by nondispatchable generators”, in Proceedings of the 9th International Conference on Probabilistic Methods Applied to Power Systems (PMAPS ‘06). [13] Koeppel, G. and M. Korpaas (2008), “Improving the network infeed accuracy of non-dispatchable generators with energy storage devices”, Electric Power Systems Research 78, pp. 2024–2036. [14] Mizani, S. and Yazdani, A. (2009), "Optimal design and operation of a grid-connected microgrid", IEEE Electrical Power & Energy Conference (EPEC), pp. 1-6. [15] Pinson, P. and G. Papaefthymiou, B. Kloeckl and J. Verboomen (2009), “Dynamic sizing of energy storage for hedging wind power forecast uncertainty”, in Proceedings of the IEEE Power Engineering Society General Meeting 2009, pp. 1-8. [16] N. Barberis Negra, J. Todorovic and T. Ackermann (2005), "Loss Evaluation of HVAC and HVDC Transmission Solutions for Large Offshore Wind Farms", Fifth International Workshop on Large-scale Integration of Wind Power and Transmission Networks for Offshore Wind Farms, Glasgow, Scotland. [17] A. Estanqueiro, (2010), "The Future Energy Mix Paradigm: How to Embed Large Amounts of Wind Generation While Preserving the Robustness and Quality of the Power Systems", Wind Power, Book edited by: S. M. Muyeen. [18] Available at http://en.wikipedia.org/wiki/Desertec, accessed on December 13. 2011

[19] International Energy Agency (IEA) (2005), “Variability of wind power and other renewables: management options and strategies”, OECD, Paris

3.3.4 DYNAMIC LINE RATING OF POWER LINES13

Electric overhead lines are subject to laws and regulations for safety and security. One requirement is that the electric conductors shall not violate clearing distances to ground or any crossing object that could come into danger if such clearing distances are smaller than those specified in National regulations or design codes. One particular requirement is that the transmission line must not carry so much electricity that the joule heating of the conductors may lead to larger sag of the conductors so that the ground clearance is violated. Normally this is defined a maximum permitted conductor temperature (thermal rating). However, the conductor temperature is also influenced by ventilation (perpendicular wind speed), solar heating, precipitation cooling, etc. Normally the line capacity (ampacity) is defined according to sets of rules which may vary with seasons. This deterministic rating is called “Book rating”. If measuring devices or weather forecasts are available with sufficient details and accuracy, the lines may be operated on a daily or hourly basis depending on actual weather. This is called “Dynamic rating”, or “Dynamic Line Rating (DLR)” and will in many cases allow for higher utilization of the transmission system.

Dynamic Line Rating of power transmission lines is therefore a method to provide more realistic calculation of line capacity. On average it allows higher transmission capacity compared to traditional and conservative static rating (book rating). While the static rating is considered as worst-case scenario (high ambient temperature, low wind speed and high solar radiation) or derived from climatology of the site, the DLR calculations are typically based on online monitoring system of the conductor temperature, sag, or weather conditions. The DLR systems supplied with meteorological measurements can be rarely seen in operational mode [1]; some attempts are made to calculate DLR climatology with simulated meteorological data [2].For operational DLR implementation, one of two methodologies is most commonly used. These are the IEEE standard [3] and the CIGRE method [4]. The two approaches were compared in a study by Schmidt [5], which found no significant differences in their results.

Normally the most critical weather parameter is the perpendicular wind speed. It is anticipated that the insulating thermal layer of air around the conductor breaks down when the perpendicular wind speed is higher than 0.6 m/s. If the average wind speed is higher, more power can be transmitted. If the wind speed is less, then the “book rating” applies. Therefore, in practice, the challenge is to predict the possibility that no single span of the line will have an average perpendicular wind speed of 0.6 m/s or less. Often this requirement will be relevant for only a few critical spans of the line.

Developed prediction DLR systems usually post-process outputs from a numerical weather prediction (NWP) model. Although the internal time step of high-resolution NWP models is usually on the order of tens of seconds, the outputs normally dumped are much coarser in time and as instantaneous values due to the amount of computing resources. The analysis of update intervals [6] of the weather data shows that averaged data are desirable and 10-minute interval is sufficient to provide accurate ampacity estimates, while longer intervals cause significant errors in ampacity determination.

The advantage of DLR starts under certain unfavorable conditions, when overhead power transmission lines are operated close to the static thermal limits. Such conditions occur now more and more frequently in industrialized

13 Lead author: Svein Fikke

countries, as the demand and generation of electricity steadily increase and more variable sources are integrated into the grid. Constructing a new power transmission line or re-conductoring an old line is very costly and DLR represents a tempting alternative. However, part of the large increase in ampacity when using 10-minute data is virtual because the utility dispatchers are not able to manage the grid with such short-time changes.

The more interesting issues occur considering weather-based renewable energy sources, such as wind turbines and solar plants. Those have been booming for the last two decades and have caused a significant decentralization of electricity production. In particular, new wind and solar farms have been built in remote areas that often lack suitable power grid connections, and the utilities often refuse to connect them to the grid or they limit their output because of their limited transmission capacity. The relation of DLR to those renewable resources is obvious when we consider input parameters of the calculation scheme – wind speed, ambient temperature and shortwave radiation. Those are exactly same variables determining production of wind and solar energy. A case study of virtual wind or solar farm and corresponding power transmission line shows limits of renewable energy production at given site [7]. Another study demonstrates how DLR can prevent or at least reduce re-dispatching of thermal power plants in areas with important wind energy production [8]. A potentially powerful application may be in connection with on-line measurements from the lines themselves. With a combination of measurements and weather forecasts the validity ranges of the measurements could be significantly increased.

References to §3.3.4:

[1] Douglas (1996): Real-time monitoring and dynamic thermal rating of power transmission circuits, IEEE Transactions of Power Delivery, 1996, 11, (3) [2] Pytlak, P., Musilek, P. (2010): An intelligent weather-based system to support optimal routing of power transmission lines, Electric Power and Energy Conference (EPEC), 2010 IEEE [3] IEEE 738-2006 (Revision of IEEE Std 738-1993): 'Standard for calculating the current-temperature of bare overhead conductors', 2007 [4] Cigre (2006): Guide for selection of weather parameters for bare overhead conductor ratings. Technical Brochure 299. [5] Schmidt (1999): Comparison between I.E.E.E. and CIGRE ampacity standards, IEEE Transactions on Power Delivery, 1999, Vol. 14, No. 4. [6] Hosek, J., Musilek, P., Lozowski, E. & Pytlak, P. (2011): Effect of time resolution of meteorological inputs on dynamic thermal rating calculations, IET Generation, Transmission and Distribution, vol. 5, no. 9, pp. 941-947. [7] Hosek J. (2011): http://presentations.copernicus.org/EMS2011-674_presentation.ppt [8] Ringelband T., Lange M., Dietrich M., Haubrich H.-J. (2009): Potential of Improved Wind Integration by Dynamic Thermal Rating of Overhead Lines, IEEE Bucharest Power Tech Conference

3.3.5 ATMOSPHERIC ICING14

Atmospheric icing on structures is a key factor when planning infrastructures, such as overhead power lines, wind turbines, meteorological stations or cable cars in arctic or mountainous regions. Detailed knowledge on frequency and duration of icing events as well as maximum ice loads are crucial parameters for the design of such structures. Insufficient investigations and knowledge on that issue may result in tower collapses combined with regional power

14 Lead author: Svein Fikke, René Cattin

outage (overhead power lines), losses in power production (wind turbines), turn-down of transport installations (cable cars) or corrupt measurement data (meteorological stations).

Atmospheric icing is defined as the accretion of ice or snow on structures which are exposed to the atmosphere. In general, two different types of atmospheric icing that impact wind turbines or overhead power lines can be distinguished: in-cloud icing (rime ice or glaze) and precipitation icing (freezing rain or drizzle, wet snow).

The different forms of atmospheric icing can be described as follows [1] and [2]: • Rime Ice: Supercooled liquid water droplets from clouds or fog are transported by the wind. When they hit a surface, they freeze immediately. If the droplets are rather small, soft rime is formed, if the droplets are bigger, hard rime is formed. Its formation is asymmetrical (often needles), usually on the windward side of a structure. Its crystalline structure is rather irregular, surface uneven, and its form resembles glazed frost. Rime ice typically forms at temperatures from 0°C down to -20°C. The most severe rime icing occurs at exposed ridges where moist air is lifted and wind speed is increased. Hard rime is opaque, usually white, ice formation which adheres firmly on surfaces making it very difficult to remove it. The density of hard rime ice ranges typically between 600 and 900 kg/m3 (ISO 12494). Soft rime is a fragile, snow-like formation consisting mainly of thin ice needles or flakes of ice. The growth of soft rime starts usually at a small point and grows triangularly into the windward direction. The density of soft rime is usually between 200 and 600 kg/m3 (ISO 12494), and it can be more easily removed. • Glaze: Glaze is caused by freezing rain, freezing drizzle or wet in-cloud icing and forms a smooth, transparent and homogenous ice layer with a strong adhesion on the structure. It usually occurs at temperatures between 0 and -6°C. Glaze is the type of ice having the highest density of around 900 kg/m3. Freezing rain or freezing drizzle occurs when warm air aloft melts snow crystals and forms rain droplets, which afterwards fall through a freezing air layer near the ground. Such temperature inversions may occur in connection with warm fronts or in valleys, where cold air may be trapped below warmer air aloft. Wet in-cloud icing occurs when the surface temperature is near 0°C. The water droplets which hit the surface do not freeze completely. A layer of liquid water forms which, due to wind and gravity, may flow around the object and freeze also on the leeward side. • Wet snow: Partly melted snow crystals with high liquid water content become sticky and are able to adhere on the surface of an object. Wet snow accretion therefore occurs when the air temperature is between 0 and +3°C. The typical density is 300 to 600 kg/m3. The wet snow will freeze when the wet snow accretion is followed by a temperature decrease.

There exists another phenomenon called sublimation which means direct phase transition from water vapour into ice, producing hoarfrost. Although it is known to cause transmission losses through corona effects, hoarfrost is of low density, adhesion and strength, and therefore does not cause significant loads on structures. Therefore it will not be considered in this chapter.

It has to be noted that in many cases the frequency of icing and the ice loads increase with increasing height above ground. This is due to a higher probability of a structure being inside clouds (icing frequency) and surrounded by high water content (ice load).

An icing event can be described with the following expressions [3], applicable to all structures and instruments exposed to atmospheric icing: • Meteorological icing: Period during which the meteorological conditions for ice accretion are favourable (active ice formation) • Instrumental icing: Period during which the ice remains at a structure and/or an instrument or a wind turbine is disturbed by ice. • Incubation time: Delay between the start of meteorological and the start of instrumental icing (dependant on the surface and the temperature of the structure)

• Recovery time: Delay between the end of meteorological and the end of instrumental icing (period during which the ice remains but is not actively formed)

Figure 21 illustrates how a wind measurement is affected by icing according to the definitions described above. When meteorological conditions for ice accretion are given (start of the meteorological icing), there is a certain delay – the incubation time - until ice accretion at the anemometer begins. By using anti-icing measures (coatings, warm surfaces etc.), the incubation time can be extended, in an ideal case until the end of meteorological icing avoiding icing of the instrument. As soon as there is ice on the sensor (start of the instrumental icing), the measurement is disturbed. Ice is accreted continuously on the sensor until the meteorological conditions for icing are not present anymore (end of the meteorological icing). But the ice will remain at the instrument for a certain time – the recovery time - until it melts or falls off (end of the instrumental icing). This delay can be much longer than the period of meteorological icing. Although the meteorological conditions for ice accretion are not present anymore, the readings of the instrument have to be discarded until the instrumental icing has ended. By using de-icing measures (heating, manual interference etc.), the recovery time can be shortened.

In order to describe the icing characteristics of a site, the following simplifications apply: • Incubation time = 0, i.e. meteorological and instrumental icing start at the same time • The duration of meteorological and instrumental icing refers to an unheated structure, typically a fully unheated anemometer of a mounting boom

Figure 22: Definition of Meteorological Icing and Instrumental Icing.

Additional parameters are needed to further describe the icing conditions at a site: • Icing rate: Ice accumulation per time [g/hour] • Maximum ice load: Maximum ice mass accreted at a structure [kg/m] • Type of ice [rime, glace, wet snow]

At sites where there is adequate solar radiation during the winter months, the ice can melt away within a rather short time after the end of the meteorological icing. At northern sites, the ice can remain on a structure for a very long time after the meteorological icing. Such site specific characteristics can be described with the Performance Index. It is defined as the ratio between instrumental icing and meteorological icing:

Instrumental Icing Performance Index = Meteorological Icing

3.3.5.1 WIND ENERGY

Icing has a strong effect on the planning and the operation of wind turbines: It influences the aerodynamics of the blades and causes production losses. Moreover, additional ice loads lead to extreme and fatigue loads. Iced wind measurement sensors at the wind turbine’s nacelle lead to erroneous behaviour and security stops. Finally, ice throw represents a significant safety risk for pedestrians and service personnel.

In most cases, wind turbines are affected by in-cloud icing, in some cases also by freezing rain. Wet snow is less of a problem for wind turbines. On wind turbines, ice usually accretes at the leading edge of the rotor blades. Icing can have the following effects on a wind energy project:

• Icing affects wind measurements and typically causes data losses and uncertain readings from the measurement equipment. • Heavy ice loads can cause the collapse of measurement towers • Ice on the wind turbine blades increases the noise levels of a wind turbine considerably • Ice throw from the blades of a wind turbine is a safety issue • Ice on a rotor blade always leads to production losses or production stop. The latter can be compared to a production ramp (no production despite good wind conditions) • Energy yield calculations for sites where icing condition prevail have a higher uncertainty compared to standard conditions. • Ice on wind turbine blades causes aerodynamic imbalance and with long exposures can increase the loading of components. • The presence of ice may make maintenance and repairs more difficult

3.3.5.2 OVERHEAD POWER LINES

Overhead power lines can be affected by in-cloud icing and freezing rain. But also wet snow often causes major problems. For electric overhead power lines, icing is often the most significant design parameter in economic terms. On electric overhead power lines, the ice accretes on a conductor in form of a cylinder causing additional loads. These additional loads lead to additional sag of the conductors which puts them outside their limits. Such additional sag can result in conductors touching each other or touching other structures such as trees and causing flashovers and short-circuits. High ice loads cause support poles, insulators and lines to break. Entire power lines can collapse and cause large power outages during ice storms. Wet snow events can also cause trees to fall on power lines and cause power breakdowns.

Another effect is the flashover at insulators of power lines leading to short circuits. It is caused generally on polluted insulators under cold fog, wet snow or ice during melting. Finally ice galloping occurs for transmission and distribution lines in many areas of the world. It occurs as the combined result of ice on the conductors and strong winds. Galloping is a low frequency, large amplitude wind-induced vibration of conductors causing flashover and large dynamic loads.

References to §3.3.5:

[1] Cigré (2006) Guidelines for meteorological icing models, statistical methods and topographical effects. Technical Brochure 291. [2] Cigré (2000) “Guidelines for field measurement of ice loadings on overhead power line conductors. Technical Brochure 179.

[3] COST 727 (2007) Atmospheric icing on structures. Measurements and data collection on icing: State of the Art. Veröffentlichung MeteoSchweiz Nr 75

3.4 WAVE ENERGY15

In order to illustrate that the above mentioned developments performed for wind and solar energy production forecasting may have potential applications in other renewable energy domains, the example of wave energy is shortly presented hereafter.

Ocean wave energy is a new, but rapidly expanding field, which offers the promise of generating significant amounts of electricity in coastal areas [1]. While only a small number of wave farms are now in operation, there are plans for significant expansion in several countries. The recently finished FP7 ORECCA project (Off-shore Renewable Energy Conversion platforms – Coordination Action) run with the purpose of creating a framework for knowledge sharing and produced a roadmap [2] for research activities in the context of offshore renewable energy with particular emphasis to wave energy which is the most mature technology with respect to tidal currents. The project indicated as a viable solution combined platforms for wind and wave energy. Status and perspectives for wave energy developments in Europe were also reviewed by [3] while a more recent global update is given in [4]. As with wind and solar, probabilistic forecasts of wave power over horizons of a few hours to a few days are required for power system operation as well as trading in electricity markets. But also as these other forms of renewable energy, wave energy can be highly variable and of limited predictability. The problem for Transmission System Operators (TSOs), utilities and trading agents is therefore to forecast ocean wave energy over horizons consistent with the operation of electricity grids and markets. These horizons are typically short, in the area of six hours in the United States or United Kingdom, and somewhat longer, in the order of two days in European electricity markets. Even though we focus here on lead times up to 48 hours, TSOs and utilities will also be interested in forecasting over somewhat longer horizons, in the range of several days, for operational planning, reserve usage and peak load matching. Very-short lead times (between 0 and 30 minutes) could also be considered in the future for the purpose of the control and optimal economic operation of the wave energy farms.

Since being a fairly new field of investigation in terms of modeling and forecasting, and since very little data exists for real-world projects of operating wave energy converters, emphasis is placed in the literature on the wave energy flux, which is the central variable representing the power these wave energy converters may harvest from the sea. The capability to forecast wave energy flux at these horizons currently exists. In existing studies, it was determined that statistical models could predict ocean wave energy more accurately at horizons of 1-6 hours, while for longer horizons, physics-based models were more accurate [5]. However, it was found that combining both methods predicted more accurately than either one individually [6], in line with parallel works that led to similar conclusions for wind [7] or solar [8] energy prediction. Finally, since the uncertainty of wave energy forecasts may be highly valuable in various decision-making problems, methodologies were proposed for the probabilistic prediction of the wave energy flux, based on ECMWF forecast of wave variables and on historical observations [9]. Overall, it is expected that a large strand of literature will follow from these works.

References to §3.4:

[1] Letcher TM: Future energy: Improved, sustainable and clean options for our planet. Amsterdam: Elsevier, 2008. [2] Jeffrey H., Sedgwick J.: ORECCA European Offshore Renewable Energy Roadmap. Published by the ORECCA Coordinated Action Project, September 2011. www.orecca.eu

15 Lead author: Pierre Pinson

[3] Clement, A, McCullen P, Falcao A, Fiorentino A, Gardner F, Hammarlund K, Lemonis G, Lewis T, Nielsen K, Petroncini S, Pontes MT, Schild P, Sjostrom BO, Sorensen HC, Thorpe T: Wave energy in Europe: current status and perspectives. Renewable and Sustainable Energy Review 6, pp. 405–431 [4] Esteban M, Leary D: Current developments and future prospects of offshore wind and ocean energy. Applied Energy 90, pp. 128-136, 2012 [5] Reikard G, Rogers WE: Forecasting ocean waves: Comparing a physics-based model with statistical models. Coastal Engineering 58, pp. 409–416, 2011 [6] Reikard G, Pinson P, Bidlot J: Forecasting ocean wave energy: A comparison of the ECMWF wave model with time-series methods. Ocean Engineering 38, pp. 1089–99, 2011 [7] Sanchez I: Short-term prediction of wind energy production. International Journal of Forecasting 22, pp. 43–56, 2006 [8] Bacher P, Madsen H, Nielsen HAa: Online short-term solar power forecasting. Solar Energy 83, pp. 1772–83, 2009 [9] Pinson P, Reikard G, Bidlot J: Probabilistic forecasting of the wave energy flux. Applied Energy 93, pp. 364- 370

4 NATIONAL ACTIVITIES

The present chapter deals with a summarized presentation of the penetration of renewable energies in the different countries which are participating to the COST Action ES1002 “WIRE”. It also gives some information about the state of the forecasting systems at the R&D level as well as in operation.

As an introduction, the following Table 4 will allow to compare for 2011 the installed capacities and the yearly electrical productions of wind, solar and hydroelectric renewable energies. The last 2 columns display the share of production of all renewable energies (wind, solar, hydroelectricity together with other forms of renewable energies such as biomass, waves…) in absolute and relative units.

Country Installed Installed Installed Production Production Production Total Renewable Renewable capacity capacity capacity 2011 Wind 2011 Solar 2011 Hydro Energies Energy share Wind Solar Hydro [TWh/PJ] [TWh/PJ] [TWh/PJ] production 2011 [%] [GW] PV/CSP [GW] [TWh/PJ] [GW] B&H 0 0 2.03 0 0 4.281 4.281 31.3 15.41 15.41 CH 0.045 0.21 13.77 0.07 0.195 33.8 37 57 0.252 0.702 121.68 133.2 CZ 0.22 1.97 2.2 0.40 2.12 2.83 7.44 8.5 1.43 7.63 10.19 26.78 DK 3.8 0.007 0.009 7.8 0.2 0.02 20.6 22.3 28.2 0.653 0.07 71.6 DE 29.1 24.8 4.4 46.5 19.0 19.5 122 20 167.40 68.40 70.20 439.20 GR 1.641 0.521 0.208 3.31 0.56 0.59 4.65 5.2 11.92 2.02 2.12 16.74 HU 0.33 0.004 0.055 0.63 0.005 0.22 2.70 6.3 2.25 0.02 0.80 9.70 IS 0 0 1.883 0 0 12.59 17.06 99.9 45.33 61.42 IT 6.86 12.75 17.95 9.56 9.26 47.67 83.12 28.7 34.42 33.,34 171.61 299.23 NO 0.51 0 30.14 1.29 0 122.1 123.4 96.3 4.65 439.6 444.2 PO 1.62 0.001 0.951 2.798 0.001 2.53 5.36 3.3 10.073 0.003 9.104 19.303

RO 1.08 16.7 17.78 0.982 0.001 6.175 3.89 0.001 60.12 64.01 26.2

4.1 AUSTRIA16

4.1.1 AUSTRIA’S RENEWABLE ENERGY POTENTIAL

In the recent past, both the electricity consumption and the generated amounts have increased. In 1990, 43.5 TWh of electricity has been consumed and 44.1 TWh produced. The percentage of renewable energies with respect to the consumption was approx. 70%. Until 2009, the consumed amount increased by about 42% and was 61.9 TWh with a share of 73.7% of renewable energy sources. The generation of green electricity has increased by 49.5% within this period (Figure 22) [1].

Figure 23: Electric power generation and power consumption – public power grid from 1990 to 2009 [1]. Source: Energie-Control GmbH

4.1.2 STATUS: WIND ENERGY

The generation of wind power has increased substantially in Austria from 2003 to 2007. At the end of 2009, 995 MW wind power were in operation. generating 1.915 GWh of electricity in 2009 [1].

The green electricity law in the version BGB I Nr. 104/2009 includes a prolongation of the expansion of wind power. The aim is an increased volume of about 1.500 GWh of power generation until 2015. The energy strategy also includes a further extension up to 1.400 MW until 2020 [1].

16 Lead author: Alexander Kann

4.1.3 STATUS: SOLAR ENERGY

Overall, the number of photovoltaic installations has increased by 59% in 2009. The energy production by photovoltaic has increased by 47% (Figure 23) [1].

Figure 24: Growth of the accepted photovoltaic installations in the Austrian states. Source: Energie-Control GmbH

4.1.4 R&D ACTIVITIES

Since 2010, the project AutRES100 [2], which is financed by the ‘Climate and Energy Funds from the Federal Government’ (Neue Energien 2020, FFG), aims at providing realistic answers to the question of cost-efficient integration of a high share of intermittent renewable energies into the power system. For this purpose a highly resolved power system investment planning and supply security optimization model is developed. With the model feasible ways to reach a 100% renewable power supply for Austria are investigated. The analyzed questions are: technological and economical provision of balancing power, power system stability, optimal adaptation of historically grown power plant portfolios, the future role of (pumped) storage options, future intelligent demand response options, grid extensions, the European interconnection system and climate change effects on the power system. On the modelling side the focus is on the detailed temporally and spatially highly resolved modelling of the variable renewable energies, of the power plant operation, of future flexible demand side options (e-mobility, heating, cooling) and of the transmission grid (DC-load flow). Investments and supply security are optimized endogenously in the model [2].

References to §4.1:

[1] Ökostrombericht, Bericht der Energie-Control GmbH gemäß §25 Abs 1 Ökostromgesetz. 2010. Austria, Vienna; available from http://www.e-control.at [accessed 09.11.2011]. [2] AutRES100: Towards 100% renewable power in Austria. Project Description, 2010. Austria, Vienna; available from http://www.eeg.tuwien.ac.at/eeg.tuwien.ac.at_pages/research/projects_detail.php?id=253 [accessed 09.11.2011].

4.2 AUSTRALIA17

4.2.1 RENEWABLE ENERGY POTENTIAL

The potential for renewable energy in Australia was extensively documented in a report by Geosciences Australia published in 2010. Titled “Australian Energy Resource Assessment”, the report considered all forms of energy including geothermal, hydro, wind, solar and ocean [1]. Australia has an excellent wind resource, and has the highest average solar radiation per square meter of any continent in the world.

4.2.2 WIND ENERGY

The installed capacity was about 2.2 GW as of 2011 [2]. Australia has a low penetration of wind power overall, around 2-3 percent. However, the state of South Australia has high wind penetration, around 20 percent on average [3], and further large wind farm developments are expected in that state. This means that Australia experiences wind farm integration challenges on a par with those in other countries with high wind penetration.

Outlook

Forecasts for wind power in Australia have estimated an additional 4 -6 GW capacity by 2020 and an additional 10 GW by 2030 [3]. The outlook for wind power is partially dependent on the realisation of solar power, because wind and solar are competing technologies in meeting Australia’s legislated renewable energy target.

Forecasting

The Australian Energy Market Operator (AEMO) implemented the European ANEMOS system which is currently meeting forecast benchmarks. AEMO have identified that in the future, with additional installed capacity, the ability to forecast wind power ramps will need to be examined more closely [3].

A CSIRO study of wind power variability showed that there are certain times when wind power variability is enhanced, often connected with cold air outbreaks over the Southern Ocean. Much of Australia’s installed wind farm capacity is affected by these large scale weather phenomena [4].

The Australian Bureau of Meteorology (BOM) is constantly improving its ACCESS model to assimilate more data, thus making it more attuned to Australian conditions than some other models.

4.2.3 SOLAR ENERGY

The majority of solar power installed in Australia is rooftop PV and there is no large scale solar generation yet. An estimated 1,450 MW of solar photovoltaic power was installed as of February 2012. In 2011 solar PV produced about 0.6% of Australia’s stationary energy [5].

17 Lead author : Aberto Troccoli

Outlook

Some rather pessimistic forecasts for solar PV have recently been substantially revised upwards. Under a moderate uptake scenario, solar PV capacity is expected to reach 5,100 MW by 2020 and 12,000 MW by 2031 [5]. However, there remains considerable uncertainty in the market and regulatory conditions.

Resource estimation

There remains a need to quantify the uncertainty in solar resource for the purpose of large scale solar generation. Satellite-based solar irradiance products are available from the BOM [6]. However, these products require further validation and testing, and surface based measurements for doing this are relatively sparse.

Forecasting

A number of research groups have been looking at the effect of solar PV spatial aggregation and are experimenting with cloud tracking algorithms for very short term forecasts. Other activities at the BOM and CSIRO include taking high frequency measurements of direct, global and diffuse irradiance components. CSIRO will use the data to model and predict solar power variability. CSIRO is also looking at the effect of spectral composition and cloud properties (e.g. height) on solar radiation. The effect of aerosols in the atmosphere may become an issue in some locations such as large cities with increasing amounts of solar PV installed.

Numerical weather prediction models are being tuned for local conditions using state of the art data assimilation methods. A new Japanese satellite (HIMAWARI-8) is planned for launch in 2014 and will increase the frequency of cloud imagery from hourly to every 10 minutes.

4.2.4 GRID MANAGEMENT

The Australian energy market operator AEMO is monitoring the issue of growing wind energy penetration. In 2011 AEMO conducted a review of international practice in wind energy integration and the technical issues that are likely to arise under growing wind energy capacity [7].

A report by CSIRO found that intermittent solar power can be managed, however good forecasts of solar power will be needed. The problem is likely to be more acute in Australia due to specific challenges such as a long and thin electrical grid [8].

4.2.5 RESEARCH ACTIVITIES

• Wind resource characterization and forecasting: UNSW and CSIRO • Solar resource characterization and forecasting: UNSW, UniSA and CSIRO • Renewable electricity grid integration (including storage): CSIRO • Large scale optimization of renewable energy systems: University of Melbourne

4.2.6 ACRONYMS

CSIRO: Commonwealth Scientific and Industrial Research Organisation UniSA: University of South Australia UNSW: University of New South Wales BOM: Bureau of Meteorology

References to §4.2:

[1] Geoscience Australia and ABARE (2010). Australian Energy Resource Assessment. http://www.ga.gov.au/energy.html [2] Clean Energy Council (2011). Clean Energy Australia Report. http://www.cleanenergycouncil.org.au/ [3] Australian Energy Market Operator (2011). National Transmission Network Development Plan. [4] Davy, R.J., Woods, M.J., Russell, C.J., and Coppin, P.A. (2010). Statistical downscaling of wind variability from meteorological fields, Boundary-Layer Meteorology 135 (1): 161-175. [5] Australian Energy Market Operator (2012). Rooftop PV Information Paper. [6] Bureau of Meteorology. http://www.bom.gov.au/climate/maps/ [7] Australian Energy Market Operator (2011). Wind Integration Investigation. http://www.aemo.com.au/Electricity/Planning/Related-Information/Wind-Integration-Investigation [8] Sayeef, S., Heslop, S., Cornforth, D., Moore, T., Percy, S., Ward, J.K., Berry, A., and Rowe D. (2012). Solar intermittency: Australia’s clean energy challenge. CSIRO Energy Transformed Flagship.

4.3 BELGIUM18

4.3.1 STATISTICS In 2009, the electricity generation in Belgium was composed of [1]:

• 53,3% of Nuclear • 40,1% of fossil fuels • 5,3% of biomass and wastes • 1,1% of wind • 0,2% of photovoltaic

At the end of 2010, the total installed capacity of wind energy was 911MW [2] and the total installed capacity of photovoltaic energy was 803MW [3]

Belgian policy targets [4] to obtain 20% of the renewable energy consumption in 2020. More precisely, the target for wind energy consumption in 2020 is focused on 9% and the one of photovoltaic energy consumption on 1%.

The forecasts of renewable electricity production are created separately by all the producers and managers of electrical grid.

4.3.2 R&D

In Walloon Region (South part of Belgium), a partnership “PREMASOL” between different university laboratories and private companies is in preparation to manage the photovoltaic electrical production.

The European project “TWENTIES” [5] creates an innovative monitoring system for overhead lines. The originality of this project is to extract the sag data directly out of the vibration frequency spectrum thanks to AMPACIMON [6] module. In a near future, this system should forecast the best load on an electric line.

Cluster TWEED [7] (Technology of Wallonia Energy, Environment and sustainable Development) aims to play a major role in the business development of «sustainable energy» sectors. Its first mission is to pave the way for the setting up of high quality and industrial-size projects in the fields of production and exploitation of sustainable energy, profitable enough to attract appropriate financial means.

“Sustainable energy” covers the following areas:

• Renewable energy sources ; • The implementation of new processes in order to achieve energy savings, energy efficiency or the reduction of greenhouse gas emissions, including CO2, at industrial level and in the tertiary sector ; • The development of products pursuing the same goals, for industry, the tertiary sector or individuals («green» services and goods).

Private company:

18 Lead author: Michel Erpicum

3E [8] is a research institution which provides technical and financial guidance in designing wind parks and solar PV systems. 3E works on many wind (off shore) and solar energy related projects:

• the development of PV and wind forecasting models • advanced offshore measurements and modeling • European PV smart grid demonstrations • etc.

References to §4.3:

[1] From the website http://www.eia.gov/countries/ , consulted online the 21 September 2011 [2] From http://www.ewea.org/fileadmin/ewea_documents/documents/statistics/EWEA_Annual_Statistics_2010.pdf consulted online the 21 September 2011 [3] From http://www.epia.org/index.php?id=18 consulted online the 21 September 2011 [4] From http://www.ewea.org/fileadmin/ewea_documents/documents/publications/reports/Pure_Power_III.pdf consulted online the 21 September 2011 [5] From http://www.twenties-project.eu consulted online the 21 September 2011 [6] From http://www.ampacimon.com/ consulted online the 21 September 2011 [7] From http://clusters.wallonie.be/tweed/en/index.html consulted online the 6 June 2012 [8] From http://www.3e.be/ consulted online the 26 September 2011

4.4 BOSNIA AND HERZEGOVINA19

4.4.1 BOSNIA AND HERZEGOVINA’S RENEWABLE ENERGY POTENTIAL

The electrical energy, produced in Bosnia and Herzegovina (B&H), comes from conventional sources. Although there is great potential in wind, solar and biomass energy, there is no such generation so far. The distribution of total electricity production is 60 % from thermal power plants and 40 % from hydro power plants (HPP). Observing the total electricity production in the last three years, there was trend of increase in electricity production: from 13.27 TWh in 2008 to 14.00 TWh in 2009 and to 15.47 TWh in 2010 [1]. A new HPP of 60 MW installed power was built and put into operation in 2010. While a significant amount of electricity produced is exported, B&H has firm intentions to take advantage of renewable energy potential and reduce CO2 emission from the fossil fuels. According to the EEDRB (Energy and Environment Data Reference Bank) the CO2 emission in B&H is about 15 million tons/year. The greatest CO2 pollution (76%) comes from coal burning (mostly thermal power plants), about 22 % from oil burning (mostly in the transportation sector) and 2 % from the natural gas burning (mostly for the domestic purposes).

The greatest potential in renewable energy (RE) sources comes from hydro potential, especially from small HPP. More than 30% of technically available hydropower capacity was exploited, so far. The major part of the electricity demand is satisfied by hydro-energy of big dams (currently 14 HPPs). Small HPP have a small share in the total demand (currently 12 small HPPs). According to the investment strategies of the energy sector in both domains, investments in small HPP will be promoted. The latest report of EBRD (European Bank for Reconstruction and Development) states that the theoretical hydro-potential in B&H is 99,256 GWh/year while the technical hydro potential is 23395 GWh/year, with 2599 GWh/year from small HPPs.

It is planned to develop 29 big HPP with total annual production of 5072.5 GWh (additional production) till 2020. In the same time faster development of small HPP is planned. The total production of these small HPP will be 960 GWh (19% of total additional production). Up to 2020 it is expected to produce 10882.4 GWh/year out which 10% will be produced by small HPPs [2].

An important potential comes from the biomass, mostly from forests and wood wastes. Forests and forestlands cover 2’709’769 hectares or over 50% of the national territory. Based on the percentage of the land area covered by forest and forestland, B&H is the forth country in Europe. The amount of forest and forestland per capita is 0.74 ha which places B&H on the sixth place in Europe [2]. Annual production of wood assortment in B&H is 6’913’783 m3 [3]. The energy potential of biomass from forestry waste (23’321 PJ) is twice higher than the potential of biomass coming from agriculture (10,164 PJ) [2]. At present time, the waste from the wood processing industry is mostly used for the household heating in rural areas (less than 50 %). These wastes are burned in disseminated, obsolete and inefficient ovens with high emission of CO2 and dust.

4.4.2 WIND ENERGY

Currently, there are no wind parks operated in B&H. There are some wind mills, in the range of kWs, installed separately, but not in significant amount.

The greatest potential in wind energy is in the southern part of B&H in Herzegovina with available sites with more than 10 m/s of annual wind speed. Wind potential measurements were performed and a wind atlas was developed for

19 Lead author: Jovan Todorovic

the entire country. This wind atlas is supposed to be a very important input for responsible Ministries in order to decide where and how wind power systems will be installed and connected to the transmission grid.

The locations with the best wind potential have a very low population density, but very weak transmission network availability. Potential investors will have to take into account additional investments in access roads and new transmission line connections. Responsible regulatory agencies and Ministries are about to reach decisions for the subsidies through the feed-in tariff and thus increase significantly the investments’ safety in wind power sources.

In parallel to the legal framework improvements, in-line with the relevant EU directives, the question of how much renewable production is acceptable without enhancement in the power system of B&H imposes itself as essential for the deployment of renewable wind energy generation. Consequently, regarding the power system topology, the necessary reserves in generation capacities and without power system upgrade, one of the speculated amounts of wind power and other RE resources to be connected to B&H power system is around 500 MW [4].

4.4.3 SOLAR ENERGY

The same obstacles exist for the solar power utilization. There are some estimations about the solar power potential in B&H: the greatest potential lies also in the southern part of B&H in Herzegovina. Theoretically, the solar power potential is about 74.65 KWh per year and the technical potential is 685 PJ. The annual average of the daily solar energy on a horizontal surface is 3.4-4.4 KWh/m2 [2]. Concerning the solar power potential, B&H is found as very perspective country. According the data available, the expected solar power ranges is 124 KWh in the north and 160 KWh in the south B&H per 1 m2 in one year [5].

There is no statistical data about operational solar energy units. Some pilot activities are established to promote solar energy utilization. The Centre for Development and Support recently concluded a project which main goal was to establish small private producers of solar installation equipment. This means that from now on solar installations are available on the B&H market [2].

4.4.4 GRID MANAGEMENT

The power system in B&H includes 400, 220 and 110 KV transmission lines and 400/x, 220/x and 110/x KV transformer stations. Connection of the RE sources in large scales will bring new challenges to the power transmission grid. Renewable sources have no dispatch nature like conventional ones, but are variable in time and space. The Transmission System Operator (TSOs) have to be able to cope with this kind of energy in order to keep the power production/consumption balance, i.e. to guarantee the power system stability. While the power system has significant reserves in transmission capacity, high penetration of the RE resources would influence the system dynamic stability. Hence, 500 MW of installed wind energy resources is the speculated limit to safeguard the power system from stability problems. But there are expectations that this amount could be higher since the power system operates transmission lines with loads far below nominal rates and is interconnected with the neighboring systems which increases its capacity to accept more wind power. It is intended to use the experience from the power systems with high renewable power penetration in order to be prepared in the best possible way to accept renewable power connections [4].

After the recently completed POWER III project on power system reconstruction, the B&H power system is in front of new challenges to include a higher penetration of renewable energy and distributed generation, to adopt new rules and to increase the communication between producers and consumers, and update it towards smart grids.

4.4.5 NATIONAL R&D ACTIVITIES

Recently, B&H has signed agreements for implementing the EC directives related to renewable energies. But the transposition of all EU directives is still on the starting blocks and the National Energy Strategy is not prepared. Consequently, there is currently no realistic program or plan for RE promotion.

Very ambitious plans are made for the development of small HPP. Till today, more than 300 concessions for small HPP development have been issued. These investment plans will be financed partly by the state owned electricity distribution companies, partly by the EBRD and WB funds and partly by foreign investors.

The wind potential is currently not used, but a number of progresses is evident especially in the southern part of the country. On the basis of collected data on the wind potential and further analysis, detailed plans for the establishment of three wind parks Borova Glava, Mesihovina and Velika Vlajna exist, with a planned total installed capacity of 128 MW with the expected gross annual production of 367.5 GWh.

The use of solar thermal panels is implemented only in limited extent in the southern part of the country, with only 4,000-6,000 m2 of solar collectors installed. The average annual production is about 3.3 GWh.

Generally, the B&H energy sector is still learning from the more experienced countries, and it is expected that the participation to the COST Action ES1002 WIRE will bring new knowledge and data for efficient use of forthcoming renewable energy sources.

References to §4.4:

[1] Annual report 2010, Power Transmission Company of B&H [Online] (http://www.elprenosbih.ba/a3/index.php?id=4) [2] ANALYSIS OF RENEWABLE ENERGY AND ITS IMPACT ON RURAL DEVELOPMENT IN BOSNIA AND HERZEGOVINA, November 2009 (http://www.euroqualityfiles.net/AgriPolicy/Report%202.2/Agripolicy%20WP2D2%20Bosnia- Herzegovina%20Final.pdf) [3] Petar Gvero, The Potential of Renewable Energy Sources in Bosnia and Herzegovina, “Climate Change in South-Eastern European Countries: Causes, Impacts, Solutions” Graz, Austria, (http://www.joanneum.at/climate/Presentations/Petar%20Gvero_The%20Potential%20of%20Renewable%20En ergy%20Sources%20in%20Bosnia%20and%20Herzegovina.pdf) [4] Jovan Todorovic, Restrictions in Power System for Renewable Production in Bosnia and Herzegovina, Ecole des Mines, Sofia Antipolis, France, March 22 to 24, 2011 [5] Association of the Renewable energy producers in B&H (http://apeorbih.com/default.wbsp?wbf_id=12&lang=loc&t=Potencijali%20OIE%20u%20BiH)

4.5 BULGARIA20

4.5.1 BULGARIA’S RENEWABLE ENERGY POTENTIAL

The Republic of Bulgaria is located in the south-eastern part of Europe on the Balkan Peninsula. The country borders with Greece and Turkey to the south, the Former Yugoslav Republic of Macedonia and Serbia to the west, Romania to the north and the Black Sea to the east. The country has a population of around 7.364 million people in 2011 and covers a territory of 110 912 (km)2.

The Bulgarian renewable energy target is for overall 16% renewable energy in 2020. Bulgaria intends to over achieve this 2020 target by 2.79%. The share of RES (Renewable Energy Source) in the gross electricity production in Bulgaria was 11.8% in 2005 (thanks to the contribution of hydropower), (EREC2020, 2010). Approximately, 3.6 billion kilowatt-hours were generated by hydroelectric power plants in 2007, (EIA, 2007).

Bulgaria has been utilizing its hydrological resources for over two centuries. In 2009, the country had 10,300 MW of installed capacity from large commercial hydroelectric power plants (HPP’s) and approximately 545 MW from small and micro (< 15 MW) HPP’s, (World Electric Power Plants Database, June 2009).

Renewables and Waste in Bulgaria in 2009, as reported by the International Energy Agency (IEA online), are also given in the following Table.

Municipal Indu Primary Biogases Liquid Geothermal Solar Hydro Solar Tide, Wind Waste* strial Solid Biofuels Thermal Photovoltaics Wave,

Wast Biofuels** Ocean e Unit GWh GWh GWh GWh GWh GWh GWh GWh GWh GWh GWh Gross Elec. 0 0 6 2 0 0 0 4053 3 0 237 Generation Unit TJ TJ TJ TJ TJ TJ TJ Gross Heat 0 0 61 0 0 0 0 Production

Bulgaria has a sizable reserve of geothermal energy and is rich in low enthalpy geothermal waters. The country has been utilizing approximately 37 percent of its total potential, or about 109.6 MW producing some 1,671.5 TJ of energy per year, for use in space heating and air conditioning, greenhouses, drinking water, and for balneology purposes (IGA, 2004). Starting in 1999, geothermal heat pumps were installed in the capital, and around 50 percent of the pumps were used for cooling. At present, there are no geothermal reserve sites that generate power (IEA, 2007).

There is good potential for utilizing biomass as an energy source in Bulgaria. While information regarding the use and potential of biomass has been limited, there have been recent developments through pilot projects and preliminary evaluations that begin to highlight Bulgaria’s full potential. In June 2008 the Council of Ministers

20 Lead author: Ekaterina Batchvarova

approved a National Long-term Programme for Encouragement of the Use of Biomass for 2008-2020. This program is a roadmap for the potential use of biomass in Bulgaria, (EIA, 2007).

The main support mechanism for RES-E (Renewable Energy Source – Electricity) in Bulgaria is a feed-in tariff in the Electricity sector, the level of which is set annually by the State Energy and Water Regulatory Commission. RES-E technologies obtain the feed-in tariff for 15 years and the scheme ends in 2015 and there is no clear view on support schemes after 2015. The New Renewable Energy Act Law (amending the Renewable and Alternative Energy Sources and Bio fuels Act), whose objective is to implement the 2009/28/EC Directive, has come in force in May 2011. The RES industry scenario is provided by the Bulgarian RES association - “Association of Producers of Ecological Energy” (APEE), (EREC2020, 2010).

The RES-E problems are related to the long lasting grid connection procedures (approximately 11 months). Therefore, many projects in Bulgaria, that have a preliminary grid connection agreement, are pending, (EREC2020, 2010).

4.5.2 WIND ENERGY

Bulgaria’s wind energy capacity has grown dramatically in recent years. Currently, the country has a wind power capacity of 86 MW, approximately 25 wind farms. A majority of this capacity was installed in 2008: the 35 MW farm, Kalchevo, and the 32 MW farm, Kavarna East. After June 2009, 14.5 MW of capacity are constructed: the Bilo (4.5 MW) and Long Man (10 MW) wind farms. Approximately 1,000 MW of capacity are planned for Bulgaria (UDI, June 2009; Nikolaev, 2007).

Resource Information

The Bulgarian wind resource is often characterized using data from published in 1982 covering the entire country and the period 1931 – 1970 (NIMH, 1982). Wind speed data from more than 160 meteorological (climate and synoptic) stations all measuring wind at 10 m above ground were used. The most promising sites for wind energy production are in the northern Black Sea Coast, the central mountain range and the Rhodopi Mountains in the south.

During the last 5-10 years, gradually, NIMH has installed automatic wind measurements at 30 synoptic stations, which form the basis for wind power assessments using contemporary software, such as WASP analysis.

Additionally, all investors organise 1-2 years observations at the sites of their interest and typically below 50 m height.

4.5.3 SOLAR ENERGY

A sizeable portion of Bulgaria’s land area receives medium levels of solar radiation. The potential for energy from this resource is greatest for low temperature thermal applications, rather than electric power generation. Warm air solar heating may be utilized in a broad range of agricultural and forestry applications such as for crop dryers and wood dryers.

Solar thermal energy has been utilized in Bulgaria in several applications. From 1977 to 1990, the Bulgarian government developed an energy efficiency program for the utilization of solar collectors, which amounted to the installation of 50,000 m2 of collectors or about 17 MW. Additional pilot and educational projects for domestic hot water heating under the PHARE program have yielded successful results, although there has not been a large increase in such projects.

Resource Information

The Sofia Energy Center, under the auspices of the FEMOPET program, estimated the total theoretical potential for solar energy in Bulgaria to be 12.955 x 109 toe. They further estimated that the technical potential for photovoltaic panels to be 53,000 toe, active thermal solar systems to be 161,000 toe and passive thermal solar energy systems to be 33,000 toe, (SEC, 2012). Bulgaria has relatively good solar potential throughout the country. The southern border of the country has its best resource.

4.5.4 GRID MANAGEMENT

Wind and solar power plant operators are currently not obliged to provide power forecasts. The development and operation of smart grids is even far from research level.

There are too many authorities involved with RES developments (5 different authorities) and the investors are not informed about the information flows between the different authorities. There is a connection fee per MW regardless of where the installation is being built. At the end, there is a selection process for which investors will be allowed to construct their plants. For this selection process, however, no criteria are defined. Regarding the grid, grid owners often deny access to the grid for projects that are licensed and already built. Moreover, Bulgaria is making no investments in smart grids and therefore the grid may have difficulties in integrating large quantities of RES, (EREC2020, 2010).

4.5.5 NATIONAL R&D ACTIVITIES

National R&D activities include both wind and solar energy topics, and are primarily related to resource estimates. Unfortunately, the responsible state institutions are not keeping contact to the research potential in country. Private developers do seek consultancy, but on small scale and limited base.

The national Institute of Meteorology and Hydrology (NIMH)is able to use for wind and solar power forecasts different mesoscale models, such as ALADIN, WRF, and MM5. Still, this can be done only within projects with additional funding.

References to §4.5:

[1] EIA, 2007 – International Energy Outlook 2007, Energy Information Administration, Office of Integrated Analysis and Forecasting, U.S. Department of Energy, Washington, DC, 20585, DOE/EIA-0484(2007), ftp://tonto.eia.doe.gov/forecasting/0484%282007%29.pdf [2] EREC2020, 2010 - European Renewable Energy Council - Mapping Renewable Energy Pathways towards 2020, EU ROADMAP, http://www.erec.org/fileadmin/erec_docs/ Documents/Publications/EREC-roadmap- V4_final.pdf [3] IEA online – International Energy Agency, http://www.iea.org/country/n_country.asp?COUNTRY_CODE= BG&Submit=Submit [4] IGA, 2004 - Mary H. Dickson and Mario Fanelli, 2004: What is Geothermal Energy? Istituto di Geoscienze e Georisorse, CNR , Pisa, Italy, February 2004, http://www.geothermal-energy.org/ 314,what_is_geothermal_energy.html [5] Nikolaev, A., 2007: Renewable energy development in Bulgaria, Renewable Energy Factsheet, European Commission - Energy and Transport, http://www.rgs.org/NR/rdonlyres/ B5DFB4AD-0C23-49E1-8F96- FF92DD8C9F92/0/BulgarianEnergyFactSheet2.pdf [6] NIMH, 1982 – Climate Workbook for Bulgaria, v. 4 Wind, Ed M. Kiuchukova, Nauka i Izkustvo, Sofia, 1982, 383 p [7] SEC, 2012 - http://www.rgs.org/NR/rdonlyres/B5DFB4AD-0C23-49E1-8F96-FF92DD8C9F92/0/ BulgarianEnergyFactSheet2.pdf [8] UDI, June 2009 - http://www.platts.com/UDIElectricityBookmark

[9] World Electric Power Plants Database, June 2009 - http://www.platts.com/Products/ worldelectricpowerplantsdatabase

4.6 CROATIA21

4.6.1 CROATIA’S RENEWABLE ENERGY POTENTIAL

Croatia is located in the southeastern Europe and central Mediterranean region with a population of about 4.5 million inhabitants. The current energy share in Croatian power system is shown in Table 5, while installed capacities for heat and electricity generation from RES in Croatia are shown in Table 6. The largest portion of energy is generated by large hydro power plants. In 2009, the share of large hydro power plants in total electricity consumption was 36.4% with only 1% contribution from small hydro, wind and biogas power plants [1]. The generation of electricity from renewable energy sources (with the exception of large hydro) is estimated to ~ 2% in 2010. Wind energy potential in the coastal and mountainous region of Croatia is important (thousands of MW), but technical potential will be generally limited by the issue of integration into the power system. Technical potential of small hydro power plants is close to 177 MW [1]. Resources for solar energy are huge along the Adriatic coast, similar to other Mediterranean countries, and mean annual solar irradiation is 1.1 – 1.6 MWhm-2 [2]. Biomass potential is estimated to 93.5 PJ, since Croatia has a traditionally active wood industry and 42% of the continental Croatia is covered by forests [3]. Finally, geothermal energy is used at several dozen places in continental Croatia, but mostly for balneology purposes. Croatia’s energy policy target is planned to reach 1200 MW of wind power by the year of 2020 to cover ~10% of the total energy consumption. According to the energy strategy, solar energy share is planned to increase 35 times by 2020, and energy from biomass is estimated to reach 68,7 PJ until 2030 [3, 4].

Table 5: Croatia’s electricity balance in 2009/2010.

Generation in Generation in Generation in Generation in Rate of change 2009 [GWh] 2010 [GWh]* 2009 [%], not 2010 [%]*, not 2009/2010[%] Energy Type incl. incl. import/export import/export

HPP 6814.4 8435.2 53.3 59.8 12.1

TPP and public 5512.5 5083.8 43.1 36.0 -16.5 CHP

WPP 54.2 139.1 0.4 1.0 132.5

Industrial CHP 395.9 446.8 3.1 3.2 2.2

Import 7580.7 6682.4

Export 1898.6 1917.4 * Preliminary data (source EIHP)

21 Lead author: Kristian Horvath

Table 6: Installed capacities for heat and electricity generation from RES in Croatia for 2010 (preliminary data).

Renewable energy Installed heat capacity [MW] Installed power capacity [MW]

Solar 64.05 0.164

Wind 0 78.95

Biomass 513.65 9.37

Small hydro 0 31.05

Geothermal 45.26 0

Total 622.96 119.53 Source: EIHP, HEP, Faculty Of Forestry University Of Zagreb, INA Naftaplin olar

4.6.2 WIND ENERGY

The present share of wind energy, produced by 7 wind power plants on the market, is close to 0.9% of the final electricity demand. However, there are over a hundred wind power projects that earned preliminary approvals, adding up to more capacity than the entire current electric power system. Wind power forecasts are carried out in collaboration between Meteorology and Hydrology Service and the national Transmission System Operator, with two components. The first one is the use of ALADIN mesoscale meteorological model for numerical weather prediction, together with simplified dynamical methods allowing for dynamical adaptation of winds to the high-resolution terrain. The second step is the statistical post-processing for the wind power forecasting itself.

4.6.3 SOLAR ENERGY

Croatia's geographical location is highly favorable for the utilization of solar energy; however, the uptake of solar energy is slow and primarily related to domestic hot water systems. Current incentives offer very high support to solar power plant developers, however with cap set to 1 MW of total PV plants in Croatia, resulting in much stronger interest in development of the wind power plants.

Due to very low penetration of photovoltaic systems, the need for solar energy forecasting is currently at the low level. The solar irradiance forecasting for general purposes is made by Meteorological and Hydrological Service. The same institution is active in satellite- and radar-based nowcasting (forecasts for 0-3 h forecast horizon) of cloud motions as well as their evolution and decay.

4.6.4 GRID MANAGEMENT

Currently, TSO uses wind power forecasting to integrate the wind power plants safely and efficiently into the system. Wind power plant operators are currently not obliged to provide wind power forecasts. Compared to wind, solar energy forecasting for Croatian TSO currently appears as a secondary priority. Development and operation of smart grids is on the research level.

4.6.5 NATIONAL R&D ACTIVITIES

National R&D activities include both wind and solar energy topics, and are primarily related to resource estimates and forecasting. Regarding the latter, wind power forecasting is the priority and enjoys sustained improvement through development of different modules of the system, on projects funded by national and international sources. The methods currently being investigated (or funding already provided for) include different formulations of meteorological models, such as ALADIN, ALADIN-DADA, WRF, WRF-LES, COAMPS, coupling of meteorological and CFD models and statistical post-processing, such as MOS and Kalman filter.

References to §4.6:

[1] Energy in Croatia, 2009, Energy Institute "Hrvoje Požar". [2] Climate Atlas of Croatia 1961.-1991., 1971.-2001., 2009, Meteorological and Hydrological Service. [3] Report "Green jobs in Croatia", 2010, United Nations Development Programme. [4] Energy Strategy of the Republic of Croatia, Official Gazette no. 130/09.

4.7 CZECH REPUBLIC22

4.7.1 THE POTENTIAL OF RENEWABLE ENERGY AND NATIONAL STRATEGIES IN THE CZECH REPUBLIC

The Czech Republic is located in Central Europe with a population of about 10.6 millions. Its energy sector is strongly based on "traditional" coal and nuclear energy. In 2011, the gross energy demand was as high as 70.52 TWh, whereas the gross production reached 87.56 TWh the same year. That made it a record year in terms of export of electrical energy. However, the power surplus has been high all the time since the Temelin nuclear plant was commissioned in 2000 – 2002. In 2011 the power production is based on coal (55 %) and nuclear (32 %), the other sources are marginal (3.5 % natural gas, 3 % biomass, 3.2 % hydro, 2.4 % solar and 0.5 % wind ) [1].

The share of renewable sources is subject to the EU obligatory renewable targets (increase to 14% of power consumption by 2020) [2], but not more. In longer-term, the support for renewables will be removed as the nuclear is mostly seen as the most preferable option for CO2 reduction.

The most authoritative source of longer-term renewable energy potential estimation is the report of "Independent energy commission" [3]. The expectations for 2050, which could be regarded as long-term potential, are given in Table 7, together with official renewable targets for 2020. The far largest potential is expected from biomass of diverse forms and usages, especially in short-term. It covers more than 50 % of 2020 renewable targets even in terms of electricity. The feasible potential of hydropower is already mostly exploited and only small increase of its usage is expected. There is a strong potential on the area of geothermal energy in the long-term under the condition that the technology using deep "hot dry rock" proves to be economical and safe. However, its use during next 10 years is expected to be limited to experimental projects. After the photovoltaic boom in 2009-2010, further PV installations have zero political support in Czech Republic. Only scarce rooftop installations are expected. However, in long-term prospects, the potential of solar power was assessed to be fairly high. Wind energy is expected to experience gradual increase and its potential is assessed as moderate.

Table 7: Czech Republic renewable energy potential

Energy Type Usage Purpose 2011 [1] 2020 [2] 2050 [3]

Electric (TWh) 0.616 1.726 18.24 Solar energy Thermal (Mtoe) - 0.022 0.19

Hydropower Electric (TWh) 2.788 2.274 2.56

Wind energy Electric (TWh) 0.336 1.496 6

22 Lead author : Emil Pelikan

100

Electric (TWh) 0 0.018 10 Geothermal energy Thermal (Mtoe) 0 0.64

Electric (TWh) 2.021 6.165 13 Biomass energy Total (Mtoe) 6.592

4.7.2 WIND ENERGY

In the Czech Republic, the technical potential of wind energy is expected to be approximately 50 TWh (23 GW) [4], of which 6 TWh (2,5 GW) was assessed as feasible [5]. The most important constraints for feasibility of wind farm development are grid connection and public opposition resulting from wind turbine visual impacts. Under the condition of increased public support for wind energy, the feasible potential was concluded to be about 15 TWh (6 GW) [6]. The wind energy potential is concentrated in highland regions of Krušné hory (along the northern border with Germany), Českomoravská vrchovina (south-central Czech Republic) and Nízký Jeseník (northeast).

The wind energy installations in the Czech Republic counted 219 MW at the end of 2011 (224 MW in July 2012 [1]), about half of it in Krušné hory. Most of these installations were constructed in 2006-2009 period, including the biggest 42 MW wind farm in late 2007. Current development is virtually paralyzed by bureaucratic obstacles. As a result, new installations in 2011 were close to zero. In spite of this, the official target of average annual increase of 50 MW wind installations until 2020 could be regarded as realistic as several medium-size projects will probably be approved and built during next years.

4.7.3 SOLAR ENERGY

The natural conditions for solar energy are similar as in Germany. Annual total solar irradiance ranges from approximately 1000 kWh/m² in the north to 1100 kWh/m² in the south of the country [7]. The average annual total sunshine duration is around 1600 – 1700 h, which disadvantages technologies focused to direct irradiance.

Until 2007, the solar energy use in the Czech Republic was limited to experimental or demonstrational projects, for instance on school buildings. However, the steep decrease of PV module prices together with slow reaction of regulatory authority, government and parliament led to excessively attractive conditions for PV projects in 2009 and 2010. As a result, there was an explosive boom of large land-based PV arrays and the capacity of PV installations came close to 2000 MW during 2010. Since 2011 the growth is slow and restricted to rooftop installations with capacity of less than 30 kW.

4.7.4 GRID MANAGEMENT

In the Czech Republic, the transmission network is operated by state-owned company ČEPS, a.s.. The distribution network is operated by private utilities ČEZ Distribuce (cca 60%), E.ON (cca 30%) and PRE (cca 10 %). The transmission network is historically robust and no blackout occurred during last decades. However, the overload caused by German circular flows during the periods of high wind power production is often presented as serious issue by network operators. Following the huge increase of applications for grid connection of PV projects, the ban

101

on further connections of wind and PV power plants was imposed in early 2010. The ban ended in September 2011 after assessment of the effects of newly installed PV capacity on the network operation. It was concluded that no emergency situations were caused by photovoltaic plants, but the power quality close to their installations decreased [8]. However the access to the distribution grid for new wind and solar installations is still severely limited by network operators.

The integration of smart grid technologies and concept of regional self-sufficiency is investigated by ČEZ Distribuce in demonstration project Smart Region Vrchlabí [9] Wind and solar play only minor role there.

4.7.5 RESEARCH ACTIVITIES IN SHORT TERM FORECASTING OF SOLAR ENERGY PRODUCTION AND DYNAMIC THERMAL RATING OF POWER LINES

The research activity is oriented to short-term solar energy production forecasting and dynamic thermal rating problems. We use the on-line weather forecasting system MEDARD [10]. The core of this system are numeric weather prediction models WRF/MM5 with a spatial resolution up to 3 x 3 km. Besides providing a weather and air quality forecast for the general public, it produces an archive of meteorological data and it serves as a testbed for development of statistical models requiring meteorological variables as inputs [11],[12].

Dynamic thermal rating of power transmission lines can provide a significant increase of transmission capacity compared to the more traditional static rating. The most important inputs to weather-based DTR systems are measured or forecast meteorological data. Recently presented case study [13] of virtual wind farm and corresponding power transmission line showed how limits of renewable energy production at given site are influenced by method of power line rating.

References to §4.7:

[1] ERÚ, Energy Regulatory Office. Annual Data Summary of Electric Power System of the Czech Republic. www.eru.cz [accessed 08.06.12]. [2] MPO, Ministry of Industry and Trade of the Czech Republic. Národní akční plán České republiky pro energii z obnovitelných zdrojů. (National action plan of the Czech Republic for the renewable energy.) 2010. http://download.mpo.cz/get/42577/47632/568798/priloha001.pdf [accessed 13.10.11] [3] Zpráva Nezávislé odborné komise pro posouzení eneregtických potřeb České republiky v dlouhém časovém horizontu. (Report of Independent expert commission for assessment of long-term energy needs of Czech Republic) http://www.vlada.cz/assets/media-centrum/aktualne/Pracovni-verze-k-oponenture.pdf [accessed 13.10.11] [4] Hanslian D. (2011): Technický potenciál větrné energie v České republice. Energetika, 61, č.8-9, 467-471. [5] Hanslian D., Pop L.: New wind map and wind energy potential of the Czech Republic. Proceedings of EWEC 2008, Brussels, 31.3-3.4 2008. [6] Hanslian D., Hošek J., Štekl J. (2008): Odhad realizovatelného potenciálu větrné energie na území České republiky. Ústav fyziky atmosféry AV ČR, Praha, 32s. http://www.ufa.cas.cz/vetrna- energie/doc/potencial_ufa.pdf [7] Photovoltaic Geographical Information System 8. (PVGIS), http://re.jrc.ec.europa.eu/pvgis/ [8] ČS RES, Czech Association of Regulated Electroenergetical Operators. Technické možnosti připojení dalších fotovoltaických a větrných elektráren do elektrizační soustavy ČR. (Technical feasibility of grid connection of further photovoltaic and wind power plants) Press Release,19.9.2011. http://www.ceps.cz/doc/soubory/20110919/CSRES_TZ_20110916.pdf [accessed 13.10.11] [9] ČEZ, a.s., Smart Region Vrchlabí. http://www.futuremotion.cz/smartgrids/cs/vrchlabi.html [10] www.medard-online.cz (15.6.2012)

102

[11] Eben K., Jurus P., Resler J., Pelikan E., Krc P.: On the impact of NWP model resolution and power source disaggregation on photovoltaic power prediction, 10th European Conference on Applications of Meteorology, 12 – 16 September 2011 Berlin, Germany [12] Brabec M., Pelikan E., Krc P., Eben K.: Musilek P.: Statistical Modelling of Energy Production by Photovoltaic Farms, Journal of Energy and Power Engineering, Vol. 5, No. 9, September 2011,pp.785-793 [13] Hosek J.: Dynamic thermal rating of power transmission lines related to renewable resources, EMS/ECAM conference 2011, Berlin

103

4.8 DENMARK23

4.8.1 DENMARK’S RENEWABLE ENERGY POTENTIAL

The main two renewable energy potentials in Denmark are wind and biomass. Due to its northerly latitude, the solar potential is technically there, but the economics of solar power are not (yet) favourable. Since the highest point of Denmark (at least the mainland – while the Farøer islands and Greenland are part of the Kingdom of Denmark, they are not part of this overview) is at only 176 m above sea level, the hydro power potential is negligible: currently exploited are 9 MW, yielding 18.9 GWh. Denmark is one of the few countries actively developing wave power, even though the resource in its own waters is good, but not stellar. The 5.5 million Danes’ electricity consumption is currently 125.2 PJ (~35 TWh). The renewable energy share of that consumption is 27%. All figures here are from the Danish Energy Agency (DEA, see ens.dk) for 2009 unless noted otherwise. It has to be noted that Denmark is to 124% self sufficient in total energy consumption, which is mainly due to the Danish North Sea oil producing 174% of the Danish oil consumption.

Figure 25: The renewable resource in Denmark, according to the Climate Commission 2010

The wind energy potential is enough to furnish the Danish electricity consumption several times over. The climate commission estimated in 2010 the potential to be 340 TWh. Denmark was also the first country to get a full wind resource map in 1999. In comparison to the usual wind atlases, the resource map is taking the micro-scale terrain into account at every location of the country. It has been verified with data from over 5000 turbines.

The solar potential with building integrated PV is estimated by the DEA to be 4-6 TWh (measured against the 33 TWh current electricity consumption). For solar thermal use, the DEA estimates that in 2030, up to 15% of all buildings collective heating demand can be furnished by solar heating (for hot water and space heating), while that figure towards 2050 could grow to 40% with higher installation rate and a shift to lower energy housing.

The potential for wave power is harder to assess, since the technology development is so diverse and in its infancy. Therefore, no standard or representative power curve exists to be the basis of a technical assessment of the resource. The gross amount of wave power passing the Danish waters every year is about 30 TWh. A number of wave power

23 Lead author: Gregor Giebel

104

devices in the Danish North Sea, some 100 km off the coast of Jutland and stretching for 150 km, could therefore produce some 5 TWh per year, according to DEA.

Figure 26: The Danish wind resource map. Source: Risø and EMD. See http://www.emd.dk/windres.

4.8.2 STATUS: WIND ENERGY

The current installation in Denmark is 3806 MW, of which 868 MW are offshore (September 2011). In 2009, wind power production was 6.72 TWh. For comparison, the total installed capacity in Denmark (including wind) was 13728 MW in 2010. The build-up on land is slow, but a new round of initiatives and the beginning repowering means that also on land new capacity is added. The larger wind farms however are being built offshore. The first ever wind farm offshore was the Vindeby wind farm north of the island of Lolland with 11 Vestas V27 which still produce electricity. The year after, the offshore wind farm of Tunø Knob was erected. The first modern large wind farm offshore was built at Horns Rev in the Danish North Sea in 2002, consisting of 80 Vestas V80 turbines for a total of 160 MW. The year after, a large offshore wind farm was built at Nysted, and a smaller one was built in the Øresund off Copenhagen on Middelgrund. Two additional farms of 200 MW each at Horns Rev and Nysted came online in 2009, and the 400 MW offshore wind farm at Anholt will be commissioned in 2012. A further project should be 600 MW at Kriegers Flak, part of a project of three wind farms in Danish, Swedish and German waters, connected to their home countries and inter-connected between each other.

105

Operators in Denmark are paid a fixed tariff, paid by Energinet.dk. This means that it is mainly the TSO who is responsible for trading of wind power.

Figure 27: The share of wind power in the grid (orange line). The green and blue bars denote installed capacity onshore and offshore, respectively. Source: Energistatistik 2010, Energistyrelsen.

4.8.3 STATUS: SOLAR ENERGY

Solar Energy in Denmark consists of solar heating and PV. Concentrating solar power is not possible due to the high amount of days with low direct solar radiation and a generally mediocre resource. In 2010, PV produced 6.1 GWh from an installed base of 7 MW. An interesting application is the use of low-temperature heat from vacuum absorbers in smaller district heating systems. The total heat from solar thermal installations was 591 TJ.

4.8.4 STATUS: GRID MANAGEMENT

Energinet.dk, the Danish TSO, was the first one to use predictions of wind power for system management in 1993 and 1994. Until 1 year ago, the grid was physically separated in a part in Western Denmark (Jutland and Fyn), which is synchronised to the larger UCTE grid in the south, and Eastern Denmark (Zealand, Lolland, Falster and some smaller islands) synchronised to the Nordic grid NordEL. Due to those well-established sizeable external connections, the Danish grid is quite stable also with high feed-in of wind power. In 2010, a 600 MW DC connection was established over the Great Belt. The main change in the Danish grid since the 1980ies was the transformation of the production sites from a few large power stations to many thousand small scale production units, connected to many levels in the Danish power system (see Figure 27 for an illustration). Wind power is the majority in numbers of those, but also distributed Combined Heat and Power plants (often fired with biomass and occasionally with a

106

large solar collector field to use solar heat) have increased strongly in number in recent years. This also means that the management of the power system is fundamentally different. A large part of it changed when the European deregulation took hold, and Nordpool, the Nordic electricity bourse, was founded. In the 1985 structure, the TSOs did rarely use their reserves, while it is now a standard occurrence.

Figure 28: The changing structure of the Danish electricity production between 1985 and 2009. Source: Energistyrelsen.

4.8.5 NATIONAL R&D ACTIVITIES

In the field of short-term prediction, Denmark has been at the forefront of the operative use and also model development. Since about 1990, there was basically continuously at least one research project funded for short-term prediction. Lars Landberg started in 1989 at Risø National Laboratory (now DTU Wind Energy) a PhD to develop a methodology which later would be known as Prediktor (see Prediktor.dk). His work was implemented on a web server at Risø in 1993 and ran for the Eastern Danish TSO, Elkraft System. Already in the year after, the Western Danish system operator Eltra got a forecasting system installed by DTU IMM (Informatics and Mathematical Modelling). This system, the WPPT (Wind Power Prediction Tool) was based on recursive statistical time series analysis and originally did not take NWP data into account. This was added to the system in 1996. Since then, WPPT has consistently performed among the best wind power forecasting tools. The success in the commercial arena was put onto a more solid footing by spinning out ENFOR from DTU, which now sells forecasting tools to clients worldwide.

From 1999-2002, the Zephyr project tried to unify the statistical (WPPT) and physical (Prediktor) approach in a single model, but failed to fully integrate both models. However, an interesting hybrid approach was described by H.Aa. Nielsen et al. [1]. The statistical power curve estimation of WPPT was initialized using the wind farm power curve from WAsP in the way Prediktor uses it. The advantage was most pronounced for the first few months of operation of the model, and for wind power classes where only few data points were available: for wind speeds above 10m/s, the NRMSE was reduced by over 30% in the first 6 months.

The Danish research collaboration between DTU and Risø (now also part of DTU) were also the first to systematically look at Ensemble forecasting, between 2002 and 2005. They (together with DMI and end users) developed a tool to forecast realistic quantiles of the distribution from both NCEP and ECMWF ensembles, and looked into the benefits of forecasting using multiple NWP input [2].

107

During another Danish funded project (PSO Intelligent Prognosis), Nielsen et al. [1] found a way to algorithmically optimize the tuning parameters for the time adaptive model, like forgetting factor and bandwidth. In the same work, they also improved the robustness of WPPT against suspicious data.

Furthermore, the two Danish partners were major partners in the three ANEMOS projects (ANEMOS, see Anemos.cma.fr, ANEMOS.plus, see Anemos-plus.eu, and SafeWind, see safewind.eu), and contributed with many improvements in physical and statistical (fully probabilistic) forecasting to the platform. During both the ANEMOS and SafeWind projects, a report on the State-of-the-Art in short-term prediction was written by Giebel et al.. While the first edition in 2003 was at 35 pages, the second edition of 2011 contains 111 pages and references over 380 papers, showing that the field of short-term prediction of wind power has been an important field of research and applications in the last decade. More detailed research results from Denmark are also included. A current research topic lies in the use of spatio-temporal correlations of wind power as input for more exact wind power forecasts downstream of some measurement points.

The Danish/German Company WEPROG (Wind Energy PROGnosis) is running a 75 member ensemble dedicated to wind energy purposes. WEPROG's Multi-Scheme Ensemble Prediction System MS-EPS (see weprog.com) has been operational since 2004. Based on WEPROGs own NWP formulation, the system is built up with three different dynamics schemes, five different condensation schemes and five different vertical diffusion schemes, which result in an ensemble of 75 members. The characteristic of the MSEPS system is that it has the capabilities to develop physical uncertainties with well-defined differences among the ensemble members. This is of advantage especially for wind energy predictions, because it means that the uncertainty is not dependent on the forecast horizon as in other ensemble approaches, but instead develops in every forecasts step as a result of the physically different formulations of the individual ensemble members (e.g. http://www.weprog.com/publications ).

Their group was also coordinator of an EU-funded project called HONEYMOON - High resOlution Numerical wind EnergY Model for On- and Offshore forecastiNg using ensemble predictions”. One part of the project was to reduce the large-scale phase errors using ensemble prediction.

Currently, the project Radar@Sea has installed a Local Area Weather Radar on the transformer platform of Horns Rev 2. The forecasts from this radar will be used for control of the wind farm.

Generally, Denmark (in comparison to many other countries, most notably UK and USA) has had continuous research funding and support for wind power and other renewable energies, which helped the build-up of world- leading institutions in the research arena (DTU Wind Energy, formerly Risø) as well as the commercial sector (Vestas as the largest wind turbine manufacturer, Siemens (formerly Bonus) as the largest offshore manufacturer, and DONG Energy as the largest offshore wind power owner and operator).

References to §4.8:

There are about 100 other publications with Danish authors, co-authors or authors from Danish institutions in Giebel et al.2011, ref [1] in the wind power section (3.1).

[1] Nielsen, H.Aa., P. Pinson, L.E. Christiansen, T.S. Nielsen, H. Madsen, J. Badger, G. Giebel, H.F. Ravn: Improvement and automation of tools for short term wind power forecasting. Scientific Proceedings of the European Wind Energy Conference and Exhibition, Milan (IT), 7-10 May 2007 [2] Gregor Giebel (ed.), Jake Badger, Lars Landberg, Henrik Aalborg Nielsen, Torben Skov Nielsen, Henrik Madsen, Kai Sattler, Henrik Feddersen, Henrik Vedel, John Tøfting, Lars Kruse, Lars Voulund: Wind Power Prediction using Ensembles. Risø-R-1527, September 2005

108

4.9 FINLAND24

4.9.1 COUNTRY’S RENEWABLE ENERGY POTENTIAL

Finland is located in the Northern Europe in between 60°N and 70°N latitude with a population of about 5.4 millions. This northern geographical location poses several challenges to the production and prediction of renewable energy. In 2010, 31% of all electricity generation in Finland was by renewable energies [1]. During the same year the total energy demand was 88 TWh, from which 77 TWh (12 % increase from earlier year) was covered by national energy production [2]. 28 % of all production was produced by nuclear power, 15% by hydropower, 36% by combined heat and power (including coal, gas, biomass and peat), and 18% by direct condensing power (coal/gas). The wind power production in Finland has been on a modest level being only 0.4 % (0.3 TWh) of all energy production in 2010.

The national energy strategy expects that largest increase in renewable energies will be provided by biomass energy production. The hydropower resource has the extra potential only about 1 TWh/yr, which makes the wind power to be the second largest source of new renewable energies in Finland, with a target of 6 TWh/yr in 2020 (2500 MW) [1]. The highest wind energy potential is located mainly on coastal regions.

4.9.2 STATUS: WIND ENERGY

The deployment of wind energy sites has been very slow in Finland. At the end of 2010, the total installed wind capacity in Finland was 197 MW producing 0.3 TWh/yr [1]. However, with the new target (6TWh/yr in 2020) and a market based feed-in tariff system, the planning of new wind energy projects has speeded up. For example, in the beginning of 2011 there were in total 5900 MW of both onshore and offshore wind power projects in various planning phases.

4.9.3 STATUS: SOLAR ENERGY

Finland's northern location poses natural challenges and limitations to the solar energy production. Nevertheless, the yearly total solar irradiance on a flat surface can reach up to 1000 kWh/m2 in Southern Finland, 900 kWh/m2 in central Finland, and even 800 kWh/m2 in Northern Finland [3]. However, there are several challenges related to solar production. Solar radiation is highly unevenly distributed during different seasons: in winter time (Dec-Jan) there are practically no solar radiation available and during mid-summer the northern parts of the Finland receive as much solar radiation as central parts of the Europe. At the moment, solar energy is mainly used by individual companies and citizens in their actions as a supplementary power source, or if they are located in the regions with no electrical network available.

4.9.4 STATUS: GRID MANAGEMENT

It is foreseen that smart electrical network represents a requirement to achieve the European Union's environmental targets. Finnish Energy Industry has the vision that Finland is acting as a part of common European level energy market, where smart grid solutions enable the electricity demand to be flexibility planned based on the available production. Therefore, the issue of smart grid production is considered important in Finland and long term plans and

24 Lead author: Sami Niemelä

109

activities have been developed. Examples of already implemented intelligent features include automatic fault locating and separation, optimization of network use and remotely read meters.

4.9.5 NATIONAL R&D ACTIVITIES

The research and development of renewable energy field in Finland is widely spread between academia, national research institutes, commercial engineering and consulting companies. The activities include e.g. wind resource mapping (and the effect of climate change) by combining measurement and modeling (NWP, CFD etc.) techniques, short-term wind power production forecasts and grid integration studies Due to our northern location, naturally one of the issues is to focus on Nordic challenges such as icing processes from the energy production point of view.

References to §4.9:

[1] IEA Wind 2010 Annual Report, ISBN 0-9786383-5-2, 2011. [2] Official Statistics of Finland (OSF): Production of electricity and heat [e-publication]. ISSN=1798-5099. 2010. Helsinki: Statistics Finland [referred: 17.11.2011]. Access method: http://www.stat.fi/til/salatuo/2010/salatuo_2010_2011-10-06_tie_001_en.html . [3] Road-map for solar energy technology and markets in Finland, Tekes-project report 594/480/00, 2001

110

4.10 FRANCE25

4.10.1 RENEWABLE ENERGY POTENTIAL IN FRANCE.

Renewable energies represent 14% in the electricity production mix in France. This number includes hydro power (11.4%). The percentages for wind and solar production are 1.4% and 0.04% respectively.

France has the second highest wind potential in Europe after U.K. The installed capacity by the end of 2010 was 5.7 GW placing it at the fourth rank behind Germany, Spain and Italy. This capacity covers around 2% of the electricity demand of the country. The objective set at national level "Grenelle de l’Environnement" for 2020 is for 19 GW installed capacity onshore and 6 GW offshore.

Regarding solar energy, by end 2010, the installed capacity exceeded 1 GWp (three times higher than the installed capacity by the end of 2009) and the objective for 2020 is 5.4 GWp.

4.10.2 STATUS OF WIND AND SOLAR ENERGY

The following two tables give an overview of the situation of wind and solar installations in France by the end of 2010. The relative numbers for Europe are also given for sake of comparison (source: Observ'ER).

25 Lead author: Georges Kariniotakis

111

4.10.3 STATUS ON GRID MANAGEMENT

Currently in France there is a feed-in tariff system for the remuneration of the production by renewable energies. Independent Power Producers do not participate directly in the electricity market. Thus, at operational level, the use of forecasts is mainly important for the French Transmission System Operator (RTE) for managing the power system. RTE has developed and operates a monitoring wind generation system (IPES) that integrates in-house models for wind power forecasting. In the future this system will be extended to include PV production. RTE together with other French actors participate in several national or European R&D projects (i.e. SafeWind, Twenties etc) on wind power forecasting and integration of renewable energies into the power system.

4.10.4 NATIONAL RD&D ACTIVITIES

In the fields of wind and solar forecasting and related remote sensing there are several research organizations and companies that are active especially in the last years. First works on wind power forecasting date back to early 90's and were carried out at MINES ParisTech. Activities on solar power forecasting are more recent although there is longer experience in the fields of solar resource evaluation. Below is given an overview of the related activities of the organisations in France that participate in the COST Action ES1002 WIRE. The French Environment and Energy Management Agency ADEME has piloted lately the development of national R&D roadmaps on smart grids, solar and wind energy. In these roadmaps wind and solar forecasting are recognised as research priorities and on this basis recent national calls for RD&D projects include these aspects.

4.10.4.1 MINES PARISTECH - CENTRE FOR ENERGY & PROCESSES (CEP)

The R&D activities in the fields of solar and wind energy at MINES ParisTech are carried out at the Centre for Energy & Processes and namely at two research Groups, the Renewable Energies and Smart grids Group (ERSEI) and the Observation, Modelling and Decision Group (OMD).

The ERSEI Group initiated research activities in the field of wind power forecasting at early 90's. Today its activities cover:

• Wind/Solar power forecasting: In this field the Group has developed several statistical and artificial intelligence based approaches, methods for probabilistic forecasting including ensembles as well as innovative prediction tools such as prediction risk indices, ramps forecasting, spatiotemporal forecasting approaches etc.

• Renewable energies integration: Several approaches have been developed for the optimal use of forecasts for decision making with applications such as strategic bidding of renewable production in electricity markets, management of virtual power plants and micro-grids, wind/storage coordination and others.

112

The ERSEI Group has participated in numerous projects for wind and PV forecasting including Care, More-Care, Dispower, Microgrids, More-Microgrids, Enseole a.o. Currently it participates in GRID4EU (FP7) for PV forecasting and has coordinated the ANEMOS (FP5), ANEMOS.plus (FP6) and SafeWind (FP7) EU projects on wind power forecasting. It is member of the team that develops and commercializes the ANEMOS wind power forecasting system (i.e. in Australia for the Market Operator AEMO).

The HelioClim project is developed by the OMD Group. This project is an initiative of MINES ParisTech / ARMINES launched in 1997, based on HelioSat algorithms. It aims at providing Surface Solar Irradiation (SSI) estimation for any site, any instant within a large geographical area and a large period of time. It covers Europe, Africa and the Atlantic Ocean. The HelioClim-1 database provides daily values of SSI, with a spatial resolution of around 30 km, for the period 1985-2005. It has been created from archives of images of the Meteosat First Generation (MFG). This database is available for free at www.soda-is.com. The HelioClim-3 database began in 2004 and has been updated daily since then. Based on Meteosat Second Generation (MSG), it delivers values of SSI every 15 min with a spatial resolution of 3 km at nadir. Like HelioClim-1, this solar database is used by customers as a long term archive of solar resource for specific geo-locations or for mapping services. It is also used, on a daily basis, for monitoring of solar power systems. Since few months, to answer some customers' needs, a special mode of HelioClim-3 has been set up to deliver, on a near real time basis, the SSI map over the field of view of MSG, every 15 minutes. This new possibility of SSI now-casting can be also used to improve monitoring applications and the quality of stochastic and/or meteorological based forecasting methods, by providing high resolution spatio-temporal information on SSI.

4.10.4.2 EDF R&D

The activities at EdF R&D focus on the following :

Wind power generation:

• Spot wind power generation forecasting (from one day to 2 weeks ahead) over France, with ECMWF and MeteoFrance data (operational system). • Participation to European projects: SafeWind, ANEMOS.plus, ANEMOS.

Solar power production :

• Short term (up to 1-2 days) and very short term (up to few hours) forecasting of photovoltaic production over La Réunion Island. Different types of approaches: regression models, classification over weather types, cam or satellite image processing. • ADEME demo project : PEGASE

Measurements :

• At EDF R&D sites (near Fontainebleau): meteo variables, PV production. • At the Island of La Reunion (meteo, camera images, …)

4.10.4.3 UNIVERSITY OF LA RÉUNION – PIMENT

PIMENT is a multi-disciplinary laboratory with 50 researchers working on areas related to physics and applied mathematics for energy and environment. The research activities tend to focus on renewable energy and smart grids.

• Research activities: The spectrum of models developed range from classical time-series model (AR, ARMA…) to Bayesian neural network model for wind and solar forecasting. An attempt to adapt models issued from financial analysis (like GARCH models) to solar forecasting is currently undertaken.

113

• Measurements & remote sensing: Measurements regarding meteorological data (and more specifically solar radiation data) and PV production are currently recorded (time-step: 1 min) at the station of Saint-Pierre at the Island of La Reunion. In addition two total sky imagers are planned to be installed on 2 distinct sites of the Island.

• R&D and Demo projects: PEGASE which is a 3-year project with EDF-SEI, EDF-R&D, Polytechnique LMD, MétéoFrance and PIMENT. The PEGASE project aims to experiment the combined use of energy storage device with renewable energy (PV and wind) production in isolated networks (like the grid of an island). Specific solar forecasting approaches will be developed and evaluated in the context of this project.

4.10.4.4 UNIVERSITY OF CORSICA

Wind speed forecasting: By studying both temporal and spatial wind speed increments from various time series gathered at different locations in Corsica and Netherlands, the research team of the Corsica University provided evidence for the intermittent nature of the mesoscale wind fluctuations. The findings suggest the existence of some "universal" cascade mechanism associated with the energy transfer between synoptic motions and turbulent microscales in the atmospheric boundary layer.

Inspired by these observations, a time series model of temporal wind dynamics associated with a continuous random cascade was built. This set-up fits remarkably the empirical wind speed distributions and allows to address the problem of wind forecasting at short-term horizons (1-12 hours ahead) leading to a systematic improvement of predictions as compared to reference models.

Solar radiation and PV production forecasting: A new approach based on Artificial Neural Networks (especially Multi-Layer Perceptron) was developed allowing forecasting solar radiation at different time-lags (h+1, d+1, h+24, m+5). Using endogenous PV production historical data and/or exogenous data (i.e. ambient temperature, direction and wind speed, nebulosity, insulation at time t), the model provides predicted signals presenting a good accuracy compared to classical references. Hybrid models (ANN, ARMA) for h+1 combining endogenous/exogenous past values and NWP ALADIN outputs from MétéoFrance have even improved these forecasting performances.

4.10.4.5 LMD – ECOLE POLYTECHNIQUE

Solar Group at LMD: The motivation of the Solar Group is to better understand and characterise the physical conditions leading to clouds and the types of meteorological regimes associated as a step for the solar forecasting and nowcasting. In particular, the research is focused on finding predicting parameters.

Research activity: Solar forecast approach from meteorological regimes, based on the daily irradiation and variability. Use of meteorological modelling and measurements, whole sky cam images, and satellite images.

Activities on Remote sensing, Measurements: Remote sensing, on-site measurements and images from hemispherical camera are performed in the SIRTA observatory (http://sirta.ipsl.polytechnique.fr/).

R&D and Demo projects: PEGASE (Prévision des Energies renouvelables et Garantie Active par le Stockage d’Energie) in collaboration with EDF-SEI, EDF-R&D, PIMENT and Météo France. The overall objective of the Group in this project is the nowcasting of the PV production over La Reunion Island

4.10.4.6 MAIA EOLIS MAIA Eolis is a wind developer and operator. The company aims at being a single partner all along a wind power project (own teams for projects development, wind farm design and expertise, engineering and construction, operation and maintenance of the turbine).

114

Research activity: MAIA Eolis carried out a benchmarking of several European forecasting services. This study is enriched with new models or services. Own forecasting model development: Functional Data Analysis (FDA) technique. Measurements, remote sensing: MAIA Eolis owns a LIDAR windcube, wind farms SCADA, sonic anemometers at 2 wind farms and supervisory control meteorological masts for some wind farms. Application: Predictions are integrated to the SCADA systems of the company. The needs of wind speed and power forecasts are therefore consequent for the management of the operation of the turbines, the optimization of the maintenance schedule and for informing the distribution and transmission systems operators about the future wind farms production. They are also used in other issues such as planning measurement campaigns (acoustic, performance...) and as a security indicator (during the building period or inspection).

115

4.11 GERMANY26

The German Federal government’s energy concept (‘Energiekonzept‘) of 2010 and the subsequent energy laws of summer 2011 presented a long-term political timetable for climate protection and the transformation of the energy supply in Germany (‘Energiewende‘). It calls for emissions of greenhouse gases in Germany to be reduced by 80% to 95% from the 1990 level by the year 2050. This transformation will result in a nearly complete shift to renewable energy sources – accompanied by a substantial increase in energy efficiency. The challenges presented by this transformation of the power system are considerable.

4.11.1 RENEWABLE ENERGY POTENTIAL

According to a recent study prepared for the Federal Government [1] Germany‘s electricity demand in 2050 could be around 500 TWh – compared to today‘s consumption of approximately 600 TWh. Many scenarios of Germany‘s future energy situation including the potential of renewable energies have been presented recently. Mostly they differ in the assumptions concerning the contribution of energy storages and in the shift between final energy carries – predominantly from chemical fuel to electricity. For the electricity sector, a complete shift to renewable energies is assumed by most of the studies.

A rather conservative assessment of the technical potential of renewable energies in 2050 as presented by one of the studies [3] is as follows: Capacity [GW] Annual energy yield [TWh]

Photovoltaics 275 250 Wind, offshore 60 180 Wind, onshore 45 180 Hydro power 5 25 Geothermal energy 6 50 Bioenergy (biogenic waste only) 25 Total 710

A recent study on the potential of German onshore wind energy [2] presented an installed capacity of 198 GW and a corresponding annual electricity production of 390 TWh showing the conservative nature of the above figures and the uncertainty in estimates of the potential. In addition, large amounts of renewable electricity may be imported from either offshore wind power sites or solar power plants in Southern Europe.

4.11.2 STATUS: RENEWABLE ELECTRICITY

Electricity generation from renewable energies increased dramatically in recent years with an annual growth rate of 17% in 2011. Currently (by the end of 2011) 122 TWh of renewable electricity is produced which is equivalent to 20,0 % of the country‘s total electricity generation (2010: 17,1%). Contributions to renewable electricity generation are by wind (38 %), bio (30 %), hydro (16 %), and photovoltaics (16 %). Geothermal energy currently only provides 19 GWh of electricity. The share of renewable energies in total final energy and total primary energy consumption are 12,2% and 10,9 %, respectively.

26 Lead author: Detlev Heinemann

116

Figure 29: German annual renewable electricity generation in GWh from hydro (blue), wind (turquoise), bioenergy (green), and photovoltaics (yellow). 1990–2011 [4].

4.11.3 STATUS: WIND POWER

In 2011, 895 wind turbines with an installed capacity of 2.007 MW have been installed – a significant increase on the 2010 figure. However, only 108 MW offshore capacity has been added. Including repowering effects, the net capacity increase is 1.885 MW. At the end of 2011 in total 22.297 wind turbines with an installed capacity of 29.1 GW were operating.

Following a year with low wind, in 2011 wind resources have been on average. This results in an increase of wind electricity from 37,8 TWh to 46,5 TWh. Wind power thus contributes 7,6 % of German electricity. Currently, a capacity of 700 MW of offshore systems is under construction.

Figure 30: German annual electricity generation from wind turbines (columns in GWh) and installed wind power capacity (solid curve in MW). 1990–2011 [4].

117

4.11.4 STATUS: SOLAR POWER (PHOTOVOLTAICS)

The fast development of the German photovoltaic market has continued in 2011 when PV systems with a capacity of 7,5 GW have been installed – nearly the same figure as in 2010. At the end of 2011 PV systems with a total capacity of 24,8 GW were producing electricity with a total energy of 19 TWh in 2011. This resulted in a share of photovoltaic energy in the German electricity generation of 3,1 % (2010: 1,9 %). For 2012 a similar increase of up to 8 GW capacity is expected.

Figure 31: German annual electricity generation from photovoltaics (columns in GWh) and installed PV capacity (solid curve in MW). 1990–2011 [4].

4.11.5 STATUS: BIOENERGY

Figure 32: German annual electricity generation from biomass in GWh. 1990–2011 [4].

Electricity generation from biomass continued to increase as well. Especially the use of biogas has increased to 17,5 TWh (2010: 14,5 TWh). In contrast, the use of vegetable oil decreased. The total sum of biogenic energies (solid and

118

liquid biomass, biogas, landfill and sewage gas, biogenic waste) amounts to 36,9 TWh in 2011 (2010: 33,9 TWh) with an installed capacity of 7,2 GW. This resulted in a contribution of bioenergy to the German electricity generation of 6,1 % (2010: 5,5 %).

4.11.6 STATUS: HYDRO POWER

Hydropower generation shows a very moderate change in installation capacity which is currently 4,4 GW. Due to the natural variability in water amount the power production decreased to 19,5 TWh in 2011

(2010: 21,0 TWh). Only small increases in capacity are expected in the forthcoming years.

Figure 33: German annual electricity generation from hydro power plants (columns in GWh) and installed hydro power capacity (solid curve in MW). 1990–2011 [4].

4.11.7 GRID MANAGEMENT OUTLOOK

As a result of the growing importance of renewable energies, new demands are being placed on electricity grids. The existing German extra high voltage electricity grid has been designed to connect large conventional power plants with areas of high electricity demand. However, the gradual concentration of wind power in North Germany, particularly in the planned offshore wind farms in the North and Baltic Sea, shows potential bottlenecks in this grid design. The successful integration of the expected amounts of offshore wind power demands a considerable expansion and restructuring of the transmission network. In addition to this new onshore grid design a specific need for an optimized offshore network to transport the wind energy from offshore wind farms to the onshore grid connection point has evolved. Although the necessity of transmission grid extensions for the transformation of the energy system has been mostly accepted, construction of new lines is often proceeding very slowly.

For photovoltaic power systems, which nearly completely feed into the low (85%) and medium (15%) voltage level of the electricity grid, the requirements for the future grid design are different. Grid stability issues due to the reverse power flow caused by, e.g., rooftop PV systems have to be addressed. To increase the number of PV systems in a certain branch of the grid the distribution grid has to be converted into a collection grid. This means the introduction of new intelligent measures to keep the grid voltage in the desired range, to provide grid services even in the case of grid failures, and to supply reactive power by inverters.

119

With large shares of renewable energies in the electricity grid, fluctuations in power generation caused by weather conditions need to be synchronized with the demand for electric energy. The residual load will show a strong increase in peak-load demand and a base-load demand dropping gradually to zero. Balancing measures are needed to minimize these fluctuations. An increasing demand for operating reserves, whose size will in future be greatly influenced by the accuracy of forecasts of the feed-in from wind and solar power. A large potential for load balancing by transporting electricity through the transmission network in Europe is expected. Additional grid transfer capacities between Germany and neighboring countries of 10–15 GW by 2030, and of 15–20 GW by 2050 are necessary [1]. Load management will also make a substantial contribution and the use of both short-term and long- term storage systems is another option for balancing. Pumped-storage units and battery systems are suitable for hourly and daily balancing; power-to-gas or the use of Scandinavian hydroelectric power in perspective for long- term balancing.

4.11.8 NATIONAL R&D ACTIVITIES

Within the topical range of the COST action, major research activities are performed in German universities, research institutes and industry. Research in offshore wind energy is largely concentrated in the research program RAVE (‘Research at alpha ventus‘) which is linked to the first German offshore test wind farm in the North Sea. In addition to engineering aspects, many meteorological topics are investigated including the use of weather information in wind power forecasting and wind farm control. Wind power forecasting has been and still is subject to several national research projects dealing with general questions of forecasting and the integration of forecasting products in the energy industry.

Corresponding national research on aspects of solar energy meteorology is less focused on a specific area. Research is mainly found in the context of grid integration of photovoltaics, but projects are performed as well linked to solar thermal power generation (direct solar irradiance modeling and forecasting) and solar forecasting in the building sector. Aside national activities, several research groups are working in the scope of international activities, and are contributing to solar power forecasting, remote sensing of solar resources, and solar irradiance studies for grid integration. They provide a significant research basis in the field of solar energy meteorology.

References to §4.11:

[1] Nitsch, J. et al. (2012): Long-term scenarios and strategies for the deployment of renewable energies in Germany in view of European and global developments. Final report for the Federal Ministry for the Environment, Nature Conservation, and Nuclear Safety (BMU). DLR Stuttgart, Fraunhofer IWES Kassel and IFNE Teltow, BMU-FKZ 03MAP146. [2] Bofinger, S. et al. (2011): Studie zum Potenzial der Windenergienutzung an Land. Fraunhofer Institut für Windenergie und Energiesystemtechnik (IWES). Bundesverband Windenergie e.V. [3] Klaus, T. et al. (2010): Energieziel 2050: 100% Strom aus erneuerbaren Quellen. Umweltbundesamt (UBA). [4] Bundesministerium für Umwelt, Naturschutz und Reaktorsicherheit (2012): Entwicklung der erneuerbaren Energien in Deutschland im Jahr 2011.

120

4.12 GREECE27

4.12.1 GREECE’S RENEWABLE ENERGY POTENTIAL

Greece is located in Southeastern Europe between Latitude 34° to 42° N, and Longitude 19° to 30° E. It covers an area of about 131’990 km² and has a population of 11’305’100 as of 2010 [1] with a density of 85.65 capita per km².

Total energy demand reached 30’629 MToe for 2010 with more than 55% in the form of oil. Total electricity installed capacity reached 12’800 MW at the end of 2010 [2] with more than 60% coming from lignite based thermal power plants, while only 15% comes from renewable energies, mainly hydro.

The vast use of fossil fuels is the main culprit for most of the emissions, which were 0.232 t/capita CO2 equivalent for fluorinated Gases, 0.610 t/capita CO2 eq. for N2O and 9.256 t/capita CO2 eq. for CO2 [3]. In order to control and mitigate GHG emissions in the country with Law 3851/2010, the Greek Government set the following target contributions of energy produced from R.E.S. in 2020:

• To the gross final energy consumption by a share of 20%. • To the gross electrical energy consumption to a share of at least 40%. • To the final energy consumption for heating and cooling to a share of at least 20%. • In transportation (biofuels) to a share of at least 10%. Furthermore the target for emission reduction is to reach 4% of 2005 levels.

Greece’s Geothermal potential, although significant, has not been exploited yet. Medium and low enthalpy geothermal fields are mostly associated with grabens and postorogenic sedimentary basins. There is a high potential in the areas of Kimolos, Polyegos, Lesvos, Chios and Samothrace islands and in the north-east continental mainland. Deep water circulation along ‘open’ faults in grabens all over the country has created a large number of low enthalpy (T=90oC) fields. There are more than 750 thermal springs with a thermal potential of low enthalpy exceeding 1000

MWth [4].

Table 8: The recent status and near future estimation of power from renewable energy sources (source: www.ypeka.gr)

Source Power from renewable energy sources in MW and percentage (%) increase

2009 2010 % 2011 (est) % 2012 (est) %

Wind 1166.9 1297.7 11 1600 23 1900-2000 19-25

Biomass 43.3 44.0 2 45 2 50-80 11-78

Hydroelectric 182.6 196.3 8 210 7 230-250 10-19

PV 53.0 198.3 274 400 102 650-750 63-88

Wind energy potential is also significant with an estimated generation potential of more than 560 TWh annually [5]. Hydro power has been the main renewable energy used in Greece for electricity production, although it is also under-

27 Lead author: Andreas Kazantzidis

121

utilized. Hydro potential is estimated at 29 TWh annually [6]. The recent status and near-future estimation of power from renewable energy sources is presented in Table 8.

4.12.2 WIND ENERGY

Greece was one of the pioneers in the utilization of wind energy. Despite the huge potential, installed capacity reached 1’208 MW at the end of 2010 [ ], quadrupling from 2001 (270 MW) [7].

As mentioned before, wind potential in Greece is more than 560 TWh annually with 305 TWh being economically competitive with today’s technology [6].

4.12.3 SOLAR ENERGY

Greece enjoys a yearly irradiance average of 1’000 to 1’800 kWh/m² on a horizontal plane with most areas being above 1’400 kWh/m² [8], making it ideal for the utilization of solar energy.

From 1976 and onwards, Greece’s main technology for the utilization of solar energy was in the form of flat-plate collectors for domestic hot water systems. Currently, Greece has an estimated installed area of more than 4’084’000 nd m² that correspond to about 2’859’000 kWth [9], putting it in the 2 place in the EU, following Germany.

Photovoltaic Systems were first installed by the Greek Public Power Company (PPC SA) in the early ‘80’s, but due to the high costs, they were limited to remote areas, mainly on the islands of the Aegean Sea. With the implementation of a new feed-in tariff scheme, the installed capacity took off during the last few years. From a cumulative installed capacity of 6.7 MWp in 2006, a new capacity of 190MWp was installed until the end of 2010 and 75MWp were added during the first trimester of 2011 [10]. It is expected that photovoltaic systems will play an important role in the future energy planning of Greece.

4.12.4 GRID MANAGEMENT

The electric grid is managed by the “HELLENIC TRANSMISSION SYSTEM OPERATOR S.A.” which was established by Presidential Decree 328/12-12-2000. The company's objective is to operate, ensure the maintenance and the development of the Electricity Transmission System across the country, as well as its interconnections with neighboring grids to ensure the country's supply of electricity in a sufficient, safe, cost effective and reliable manner.

4.12.5 A NATIONAL PROJECT: THE HELLENIC NETWORK OF SOLAR ENERGY (HNSE)

The newly built Hellenic Network of Solar Energy (HNSE) aims to investigate the level of available solar energy over Greece with high spatial and temporal resolution. It consists of a satellite image analysis algorithm for the derivation of cloud properties and the calculation of solar irradiance since 2003, a ground–based network for validation of results and a prediction tool of incident solar energy based on meteorological forecasts. These tools have been separately developed in previous years. During the pilot phase of the project, common protocols and databases are developed and comparison of results performed.

The cloud estimation on solar radiation is based on satellite Meteosat Second Generation (MSG) images with 0.05o and 15min spatial and temporal resolution respectively. Model calculations are used for the estimation of solar irradiance at the ground for cloud-free and cloudy skies. The ground-based meteorological network consists of 15 stations across Greece equipped with Kipp&Zonen CMP11 and CM11 pyranometers. The central calibration facility is located at the University of Patras.

122

Forecasts of solar irradiance are provided from a WRF model with 1-h steps and 2, resp. 10 km spatial resolution. The model validation with the satellite derived estimates and the ground-based measurements are expected to be completed in 2012.

References to §4.12:

[1] Eurostat Demography Report 2010 [2] Eurostat 2010 [3] http://www.eea.europa.eu/data-and-maps/data/national-emissions-reported-to-the-unfccc-and-to-the-eu- greenhouse-gas-monitoring-mechanism-5 (accessed 2/9/2011) [4] M. Fyticas, A. Arvanitis, “Geothermal potential in south-east Europe”, 3rd South East Europe Energy Dialogue, 2009. [5] B. Lehner, G. Czisch, S.Vassolo, “EuroWasser: Europe’s hydropower potential today and in the future”, Kassel, 2001. [6] EEA Technical report, “Europe's onshore and offshore wind energy potential - An assessment of environmental and economic constraints”, No 6/2009. [7] Ministry of Development, “Energy Outlook of Greece”, Athens, 2009. [8] www.re.jrc.ec.europa.eu/pvgis (accessed 10/10/2011) [9] European Solar Thermal Industry Federation, Solar Thermal Markets in Europe 2010, June 2011. [10] Hellenic Association of Photovoltaic Companies, “Greek PV Markets Statistics” 2010-11, 2011.

123

4.13 HUNGARY28

Hungary is located in the Carpathian Basin surrounded by the Alps and the Carpathians. In the Basin there are mostly plain terrains and only some hilly and mountainous regions contribute to the topography variety.

After 1990 some previously blooming industries started deteriorating, and other branches could gain ground. The whole industry changed its profile during the 90’s and as a consequence CO2 emission dropped rapidly. Hungary ratified the Kyoto Protocol in 2002 which declares that besides reducing CO2 emission other requirements have to be achieved including the rate of green energy.

4.13.1 POTENTIAL FOR RENEWABLE ENERGY

Hungary has large potentials in biogas, biomass, geothermal and solar energy, but due to its location in a basin wind energy is moderately available. In the previous decade green energy production grew rapidly. According to the Renewable Energy Strategy of the Government, the total feasible renewable energy potential can exceed 2600 PJ/year (Table 9).

Table 9: The estimated achievable renewable energy potential of Hungary (Source: Hungarian Academy of Sciences, Renewable Energy Subcommittee, 2009)

Renewable Energy Potential (PJ)

Solar photovoltaic 1750

Biomass 300

Solar thermal 102.5

Geothermal 63.5

Water 14.4

Wind 532.8

Total 2600-2700

4.13.2 WIND ENERGY

Wind speeds are moderate over Hungary. The annual average wind speed at 10 m.a.g is around 2-4.5 m/s. The highest wind speeds occur over northern Transdanubia where the average wind speed at75 m.a.g is above 5 m/s, and the estimated annual mean potential is 180W/m2.

28 Lead author: Katalin Toth

124

Figure 34: Average wind speed at 75 m.a.g. over Hungary (Source: Hungarian Meteorological Service, 2005)

The efficiency of wind energy production is supported by competitive feed-in tariffs (around 0.11 EUR/kWh) and guaranteed quotas. The maximum of currently grid capacity is 330 MW, while the installed wind power capacity at the end of 2010 was 295 MW.

4.13.3 SOLAR ENERGY

Solar radiation measurements at the Hungarian Meteorological Service (HMS) have long tradition. This fact is well- represented by our total global radiation data series for Budapest (Marczell György Main Observatory of the HMS). Measurements started in the thirties of the previous century. Due to this fact it is one of the longest total global radiation data series in Europe.

The areal distribution of yearly totals of sunshine duration and of the total global irradiance for Hungary can be seen in Figure 34 and Figure 35. Concerning sunshine duration, the formation of isolines is determined by the latitude and continental characteristics. As a consequence, yearly sunshine duration characteristically reaches its maximum values of around 2000 hours in the southern region of the country. Similar pattern was found in the case of total global radiation, but some differences can be seen: the highest yearly totals can be found in the south-eastern part of Hungary where the values approach 5000 MJ/m2.

Figure 35: A real distribution of yearly total of sunshine duration in Hungary (Source: Hungarian Meteorological Service, 2010)

125

Figure 36: Areal distribution of yearly totals of total global radiation in Hungary (Source: Hungarian Meteorological Service, 2010)

4.13.4 THE NATIONAL CONTRIBUTION TO COST ACTION ES1002 WIRE

As mentioned before, the highest wind speeds can be found in the north-western part of Hungary (Figure 33). Although, most of the wind power stations are located in this area, many power stations have been installed in other parts of the Carpathian Basin. The Hungarian Meteorological Service provides wind forecasts for a number of power stations across the country. For doing this, the HMS has been using for about 9 years a special configuration of ALADIN LAM model called dynamical adaptation of the wind field. A short (30 min) adiabatic forecast is running in order to get the wind field adapted to a higher resolution orography (5km) over the Hungarian territory. Furthermore, a new non-hydrostatic, high-resolution operational model has been acquired at the beginning of 2010 (AROME model) which is able to substitute the dynamical downscaling. The comparison of the two different model results will be the subject of our investigation with the observational data of the wind power stations involved in the research.

As for the electrical grid system, keeping the maintenance on the lowest cost is a central issue. Weather can affect the energy transportation in many ways. In winter freezing precipitation can mean potential danger to the transmission lines: if the loads reach a critical mass, overhead wires break down and the reconstruction is prolonged and costly. Therefore, understandably, the forecast of these phenomena is essential for the energy companies: the main goal is to involve numerical weather prediction models into the forecasting process and to use ensemble methods for probabilistic predictions.

126

4.14 ICELAND29

4.14.1 ICELAND'S RENEWABLE ENERGY POTENTIAL

The primary energy generation in Iceland is a mixture of hydro- and geothermal energy production which in 2010 represented 95% of total electricity production of nearly 12.592 and 4.465 GWh, respectively. Carbon fuels produced less than 5% of the total installed capacity. Approx. 80% of the primary energy use in Iceland comes from hydro- and geothermal power [1]. Due to the orography and the location of Iceland in the North Atlantic storm track, Iceland has large wind energy potential. Additionally, there is potential to produce electricity with other methods such as wave energy and biomass.

Figure 37: Primary energy use in Iceland 1940-2010 [1].

4.14.2 STATUS: WIND ENERGY

As of today virtually no electricity is produced by wind turbines and fed to the power transmission grid. However, the potential is being actively investigated and both private and governmental companies have shown a large interest.

29 Lead author : Hálfdán Ágústsson

127

Recently Landsvirkjun (The National Power Company in Iceland) has announced that it will erect two wind turbines (2 MW) in South-Iceland for a further research of the wind energy potential.

4.14.3 STATUS: SOLAR ENERGY

Utilization of solar energy is mostly limited to the use of photovoltaic systems powering small systems in remote areas, both private and governmental, i.e. lighting in isolated huts, weather stations and other low voltage technical equipment. A significant increase in the utilization of solar energy is not expected.

4.14.4 STATUS: GRID MANAGEMENT

Grid management in Iceland is the responsibility of Landsnet which was founded in 2003 to operate Iceland’s electricity transmission grid. The grid management is strongly influenced by the location of hydro- and geothermal stations, orography and natural hazards, such as weather, earthquakes, volcanism, avalanches and atmospheric icing [2]. Steps have been taken to prepare the grid for the connection to future wind energy farms. The Icelandic electric grid is not connected to the European grid but the feasibility of an undersea cable has been discussed.

4.14.5 NATIONAL R&D ACTIVITIES

The weather 1957-2010 in Iceland has been dynamically downscaled to high horizontal resolution and the dataset, as well as observations from a dense network of automatic weather stations, are used for research related to for example severe weather, wind energy and atmospheric icing. The wind energy potential is actively being investigated, e.g. in the framework of the Icewind project and in other projects by both private and governmental institutions. Landsvirkjun intends to erect wind turbines at a pilot site in Iceland but various regions in Iceland show significant wind energy potential. Methods to simulate and operationally forecast atmospheric icing are being investigated [3] and a wind- and icing-atlas is in preparation based on the aforementioned datasets. Furthermore, a system is being developed where atmospheric observations from a small and cheap autonomous aircraft are used to improve high- resolution atmospheric simulations as well as operational or on-demand forecasts of local weather in complex orography, e.g. for wind energy applications [4].

References to § 4.14:

[1] Energy statistics in Iceland 2011, The Icelandic Energy Authority, http://os.is, 6. June 2012. [2] Arni Jon Eliassson, 2005: Natural hazards and the Icelandic power transmission grid. Cigre conference, Slovenia, Velenje 2005. [3] Arni Jon Eliasson, Egill Thorsteins, Halfdan Agustsson and Olafur Rögnvaldsson, 2011: Comparison between simulations and measurements of in-cloud icing in test spans. IWAIS, international workshop on atmospheric icing on structures, China, 2011. [4] Halfdan Agustsson, Haraldur Olafsson, Marius O. Jonassen and Olafur Rögnvaldsson, 2011: Improving high- resolution simulations of local weather using observations from a small unmanned aircraft. ICAM, international conference on alpine meteorology, Aviemore, Scotland, 2011.

128

4.15 ISRAEL30

4.15.1 ISRAEL’S RENEWABLE ENERGY POTENTIAL

The State of Israel is located in the Eastern Mediterranean region with a population of approximately 7.4 million inhabitants. Israel has undergone rapid economic development during the past 10-15 years, and has reached the standard of living typical of many countries in western and southern Europe [1].

The Israeli electricity system is almost totally reliant on fossil fuels, mostly coal and natural gas [2, 3]. The share of renewable energy (RE) in power generation is currently negligible, approximately 0.44% of national electricity production [4]. Israel has pioneered several developments in the field of alternative energy, including flat solar collectors for domestic use, solar ponds, and parabolic trough technology [5].

Despite Israel’s proven capacity in RE research and development, many promising RE developments have remained at the research stage [1]. This is because of the lack of resources and policy coordination necessary simply make the initial assessment of their commercial viability. The situation is presently undergoing positive changes mainly because it is now well known that the utilization of renewable energy resources is one of the most efficient ways to reduce anthropogenic emissions. Recently, on July 17, 2011, the Israeli Cabinet approved a plan to support the production of renewable energy, seeking 10% of electricity production with such methods by 2020 [6]. The aim of the planification is to reduce emissions and pollution in the electricity sector. The Cabinet set a target of 1’550 megawatts (MW) of electricity from renewable resources by the end of 2014, and 2’760 MW by the end of 2020 (Table 10) [7].

Renewable energy technology Predicted energy production [MW]

2014-2015 2016-2017 2018-2019 2020

Wind energy 250 400 600 800

Biogas and biomass energy 50 100 160 210

Large (PV and TS) solar energy systems 700 750 1000 1200

Middle PV solar energy systems 350 350 350 350

Small rooftop PV energy systems up to 50KW 200 200 200 200

30 Lead author: Pavel Kishcha

129

Total RE production 1550 1800 2310 2760

Percentage of RE production in the total electricity production 5.3% 6.5% 8.3% 10.2%

4.15.2 SOLAR ENERGY

Like most of its eastern Mediterranean neighbors, Israel has one of the highest solar radiation rate in the world. At present, solar heating is mainly used for domestic heating in most buildings. Regulations in place since 1980 require the installation of solar water heaters in all new buildings. As a result, rooftop flat-plate solar collectors supply today domestic hot water to 75% of the households, making Israel a world leader in solar water heating [3].

Solar energy research is conducted both by the Israel Electric Corporation and by Universities, especially the Weizmann Institute of Science, Ben-Gurion University, and the Israel Institute of Technology (Technion) [3, 5]. The Ben-Gurion National Solar Research Center is involved in both thermal solar and photovoltaic applications and is used as a testing ground for a variety of demonstration facilities. The Solar Research Facilities Unit at the Weizmann Institute of Science focuses on the utilization of concentrated solar energy. Research projects at the Weizmann Institute include the development of advanced technologies for high-temperature heat and for electricity generation, storage and transport of energy [3]. Multidisciplinary scientists at Technion are pioneers in the field of nano-energy: they have discovered that nano-sized materials consisting of nanocrystal quantum dots can absorb sunlight not only in the visible range, as materials currently used in solar panels do, but also in the infrared and UV ranges. This makes them ideal for photovoltaic cells used to turn sunlight into electricity, promising much more efficient solar power conversion [5].

By the year 2020, quotas will be set for obtaining 1200 MW from large solar energy systems, 350 MW from middle solar systems, and 200 MW from small rooftop solar systems up to 50 KW (Table 9). The developing solar energy industry needs extensive territories to build large solar plants. This may be a problem for Israel which has limited territory. Nevertheless, an extensive land area of approximately 350,000 dunams in the Negev desert (South Israel) has been assigned for the installation of solar power plants [7].

4.15.3 WIND ENERGY

Israel’s wind potential is rather low and faces the additional difficulties of location and grid interconnection [1]. Israel currently operates one wind farm with an installed capacity of 6 MW and is completing preparations for the operation of two more farms with a 50 MW capacity. Wind energy potential is limited to those areas with sufficiently constant wind, some of which are being opposed by green groups on landscape conservation grounds [3]. Nevertheless, in accordance with the aforementioned decision of the Israeli Cabinet, a quota of 800 MW has been set for wind energy in national electricity production by the year 2020 (Table 9). A number of wind farms are planned in the Galilee, the Arava and the Negev [7].

4.15.4 BIOGAS AND BIOMASS ENERGY

Another potential source of renewable energy is the biogas produced by anaerobic digestion processes in the municipal, industrial and agricultural sectors. The Israeli Ministry of Environment Protection has commissioned surveys to assess the energy potential of methane-rich biogas, the potential of anaerobic processes in municipal wastewater treatment plants and the industrial treatment of organic wastes and agricultural treatment of poultry, dairy, cattle and other agricultural manures and waste [3]. Trials have been made for recycling organic solid waste

130

produced in small Bedouin villages in the Negev, for their benefit [8]. A quota of 210 MW will be set for obtaining biogas and biomass energy in the national electricity production by the year 2020 (Table 9).

4.15.5 GRID MANAGEMENT

Existing electric grid technologies can only allow the input of 5-20% of electric energy from renewable energy sources into the national electric grid [9]. This is because of its inability to tolerate fluctuations. Smart-grid technologies are now under development in order to devise a way to incorporate more renewable energy sources into the electric grid than is currently possible with conventional technologies [9].

References to §4.15:

[1] Mor, A., and Seroussi, S., 2007. Energy efficiency and renewable energy Israel - national study. Plan Bleu. Sophia Antipolis, March 2007. (www.planbleu.org/publications/atelier_energie/IL_National_Study_Final.pdf ). [2] Worldwide electricity production from renewable energy sources. 12th inventory, Observ'ER (the industry reference for renewable energies), 2010. ( http://www.energies-renouvelables.org/observ- er/html/inventaire/Eng/preface.asp ) [3] Renewable energy. The official information from the Israel Ministry of Environment Protection’ homepage. (http://www.sviva.gov.il/Enviroment/bin/en.jsp?enPage=e_BlankPage&enDisplay=view&enDispWhat=Zone& enDispWho=renew_energy&enZone=renew_energy) [4] http://en.wikipedia.org/wiki/List_of_countries_by_renewable_electricity_production#cite_note-25 [5] Solar power in Israel. Wikipedia. (http://en.wikipedia.org/wiki/Solar_power_in_Israel#Dead_Sea_Solar_pond) [6] Reuters (Tel-Aviv), July 17, 2011. Israel eyes 10 percent of energy from renewable by 2020. (http://www.reuters.com/article/2011/07/17/us-israel-energy-idUSTRE76G0WQ20110717). [7] The National Infrastructure Ministry's policy of embedding renewable energy sources into the Israeli electricity system, 2010 (http://www.mni.gov.il/NR/rdonlyres/D2E4FE39-80D6-4E05-AC44- E9C0BCDE5FEB/0/renewables.pdf) [8] Developing Integrated Waste Recycling Systems for Agricultural and Environmental Safety using Biogas Digesters in Unrecognized Villages for the Health Promotion of Bedouin Women. (http://web.bgu.ac.il/NR/rdonlyres/718C7DB8-8BD6-47CD-9C2A-16E824166341/118732/42009.pdf) [9] Smart-Grid Innovation in the Eilat Region, Israel. (http://renewable-energy-eilat.org/innovation/smart-grid/

131

4.16 ITALY31

4.16.1 ENERGY DEMAND IN ITALY, (ACTUAL & GROWTH)

The annual energy demand in Italy in 2010 has been of 332 TWh and since 2006 Italy has become a summer peaking country with the largest annual peaks reached in June-July. In particular, in 2011 the annual peak of 56.5 GW has been reached in July. In the Grid Development Plan by TERNA, the Italian TSO, a Compound Annual Growth Rate (CAGR) equal to +2.3% is assumed for the period 2009-2020 for the “high” scenario corresponding to a total demand of 410 TWh in 2020. For 2015 TERNA assumes an electric demand equal to 362.1 TWh with a CAGR equal to +2.1% in 2009-2015 and +2.5% in 2015-2020. TERNA estimates the peak load in 2020 as 74 GW.

4.16.2 SOLAR AND WIND ENERGY POTENTIAL (ACTUAL & DEVELOPMENT)

Italy is the third European Country for the installed wind power, after Germany and Spain and sixth in the world32.

The areas with the highest wind potential are located in Central/South Italy and in the major islands. The amount of installed wind power up to 2011 is 6860 MW and it is expected to rise up to 9600 MW within 2013/2014. ANEV – the Italian National Association on Wind Energy – forecasts a total installed wind power in 2020 equal to 16 GW, most of which (around 15 GW) are concentrated in the Central/South Italy.

As for wind, solar potential is mainly located in South Italy, but a significant part is installed in the North too. Solar energy in Italy comes from mainly PV plants, for a total of about 12.750 MW. The only thermal CSP project in Italy is a 5 MW power plant with molted salt as fuel and storage medium still in experimental phase and located in Priolo (Sicily).

Table 11: Net electricity production in Italy (Source: Terna and GSE)

% wind + Thermoelectric % RE vs Wind Total Total PV vs TWh Biomass PV geo hydro total (gas, oil, coal, power RE production total production waste not bio) production

2010 211.7 9.3 9.0 1.9 5.0 53.8 79.0 290.7 27.2 3.7

2011 206.5 11.3 9.6 9.3 5.3 47.7 83.1 289.2 28.7 6.5

Table 11 shows the energy production from Renewable Energy Sources RES (compared to the total energy production) in 2010 and 2011. It can be noticed that the main RES type remains hydroelectric energy while the biggest annual relative increment is due to wind and solar energy.

31 Lead authors: Paolo Bonelli, Anna Maria Sempreviva, Andrea Pitto

32Global Wind Energy Council

132

Table 12 shows the net power installed in Italy. The wind and solar installed power has increased significantly from 2010 to 2011, confirming the positive trend of the last years.

Table 12: Net Power in Italy (Source: Terna and GSE)

Thermoelectric % wind % RE vs (gas, oil, coal, Wind Total Total + PV vs GW PV Geo hydro total waste) power RE power total power power

2010 75.0 5.8 3.5 0.73 21.5 31.5 106.5 30 8.7

2011 75.0 (2010) 6.9 12.8 0.77 18.0 38.5 113.5 33.9 17.3

The minimum 2020 targets of national Action Plan are 12.7 and 8.6 GW of installed power for wind and solar respectively, including 600 MW from concentrating solar plants, and 680 MW from offshore wind farms. The corresponding annual energy targets in 2020 are 20.0 and 11.4 TWh from wind and solar plants respectively.

More than 50% of wind installed power is shared among 7 different producers, with IP-Italia and ENEL Green Power in leading position.

4.16.3 PUBLIC BODIES INVOLVED IN THE ELECTRIC GRID, RES MANAGEMENT AND RESEARCH

Terna SpA (www.terna.it)

Terna is a leading grid operator for energy transmission. With over 63,500 km of lines, the company is the owner and operator of the National HV Electricity Transmission Grid. Terna safely transmits HV electricity from production centres to consumption areas throughout Italy. Some relevant aspects related to RES integration in the Italian grid are:

• Possible grid congestions on sub-transmission level (150/132 kV); • In the Italian energy market, RES have priority in dispatching at the same offer price; • TERNA often limits the power output from wind parks for grid security reasons; • Compensation mechanisms for missed production and the replacement of green certificates for small wind farms; • The new remuneration mechanism, introduced in 2010, for missed wind power productions, is no longer based on historical data on wind farm productions, but on the reconstruction of the missed available production on the basis of wind measurements. • After 2007 TERNA adopted a tool for short-term forecasting of wind power, based on the processing of power measurements at “reference” wind farms via neural networks. In 2010 the deliberation of the Italian Authority invited TERNA to continue developing and improving this tool, which allows to reduce balancing costs and to assure the availability of sufficient reserve capacity at low costs. Moreover, the same deliberation imposes that the missed wind production due to TERNA’s modulation actions should be inserted into the energy volumes on

133

the dispatching service market. Such mechanism incentivises TERNA to reduce the modulation actions on the wind power plants, by first modulating the other generating units. • Photovoltaic forecast is also important because, although the majority of PV plants is installed in the low and medium voltage system, the big amount of installed capacity (more than 11 GW) significantly affects the “equivalent” load “seen” at the transmission level hence the generation dispatch, cost and security of the system.

GSE SpA, GestoredeiServiziEnergetici, (www.gse.it ) plays a central role in promotion, support and development of renewable energy sources in Italy. GSE’s sole shareholder is the Italian Ministry of Economy and Finance which, in consultation with the Ministry of Economic Development, provides guidance on GSE’s activities. The main activities related to RES forecasts are: • Market trading of electricity generated by small producers of RES; • Running the net metering service (“scambio sul posto”) for electricity generated by RES plants of up to 200 kW or high-efficiency CHP plants.

AEEG, Autorità per l’Energia Elettrica ed il Gas, (www.autorita.energia.it). The Regulatory Authority for Electricity and Gas is the independent body which regulates, controls and monitors the electricity and gas sectors and markets in Italy. The Authority's role and purpose is to protect the interests of users and consumers, promote competition and ensure efficient, cost-effective and profitable nationwide services with satisfactory quality levels. Its mission includes defining and maintaining a reliable and transparent tariff system, reconciling the economic goals of operators with general social objectives, and promoting environmental protection and the efficient use of resources.

RSE SpA, Ricerca Sistema Energetico (www.rse-web.it). RSE carries out research into the field of electrical energy with special focus on national strategic projects funded through the Government Fund for Research into Electrical Systems. RSE is a total publicly-controlled company: the sole shareholder is GSE. The activity covers the entire supply system with an application-oriented, experimental and system-based approach. About RES, a group in RSE works in the field of interaction between Atmosphere and the Electric Systems. The group activities are:

• Application of meteorological modelling to the assessment of RES capability, both solar and wind; • Meteorological modelling and local post-processing applied to wind and solar irradiance, both global and direct; • Climatic change and their impacts on the electro-energy system; • Running a meteorological station at Milano and collecting data from other station in different Italian sites; • Environmental impact and acceptance of RES.

CNR, Consiglio Nazionale delle Ricerche www.cnr.it. CNR is the largest Italian public research body with 109 Institutes grouped in 11 Departments, under the Ministry of Education, University and Research (MIUR). CNR has been amongst the promoters of the newly formed Italian Alliance for Energy Research AIREn together with ENEA and the major Italian Polytechnic Universities. The Energy and Transport, Earth and Environmental Sciences and Production System are the departments most involved in RES research including smart grids, modeling for RE siting and resource assessments. Concerning the COST Action ES1002 WIRE, the main institute involved is the Institute of Atmospheric Sciences and Climate (ISAC: www.isac.cnr.it) coordinator for the CNR participation in the European Energy Research Alliance EERA Joint Programme Wind Energy. CNR and in particular ISAC own several unique and large experimental facilities in different environments spread around Italy including coastal, offshore, mountainous and in extreme climate (Arctic and Antarctic regions and Himalaya) allowing developing and testing in house instruments including ground- based remote sensing instruments as well as several atmospheric models from global to microscale for weather forecast and process studies. Added values of new remote sensing techniques both ground –based and space-borne have been used to improve short-term forecast of wind and solar capacity.

ENEA, Agenzia nazionale per le nuove tecnologie, l’energia e lo sviluppo economico sostenibile, (www.enea.it ). Concerning RE, ENEA is involved in research about technologies for the transformation of solar and wind energy

134

in electricity and heat. It runs some monitoring solar stations, as Casaccia, near Rome, and Lampedusa isle. ENEA and RSE run a web site that provides direct normal irradiance forecast on some sites in Italy.

References to §4.16:

[1] Terna – Electric Energy Balance, 2009, 2010- http://www.terna.it/default/Home/SISTEMA_ELETTRICO/statistiche/bilanci_energia_elettrica/bilanci_nazion ali.aspx [2] Terna – www.terna.it [3] GSE – www.gse.it [4] RSE – www.rse-web.it [5] ENEA – www.enea.it [6] CNR – www.cnr.it

135

4.17 THE NETHERLANDS33

4.17.1 COUNTRY’S RENEWABLE ENERGY POTENTIAL

The production of electricity through renewable energy or ‘green electricity’ is growing slowly, but continuously in The Netherlands. In 2010 the share of green electricity increased to 9% of the total consumption (in 2009: 8.9%, 2008: 7.5% and 2007: 6%) The increase is a consequence of more installed wind turbines and more use of biomass. Compared to many other European countries the Netherlands does not have much renewable energy; this is partly due to the fact that the Dutch Government‘s support for renewable energy developments has been lower than in other European countries like Spain, Denmark and Germany.

4.17.2 STATUS: WIND ENERGY

Wind energy production in The Netherlands decreased during 2010. This is compensated by the expansion of biomass. The Dutch government has subsidies for building new wind farms as part of the “Regeling Duurzame Energieproductie” (SDE). Based on the subsidies, wind farms totaling 1500 MW could be built. In 2010 this was 70 MW (subsidy based). 2010 was a bad wind year; the production of wind energy was about 23% below expectation. Most wind farms in The Netherlands are built in the coastal area. Many requests for subsidy to build new wind farms were submitted and are under evaluation by the Dutch Government. In 2020 the share of wind energy in the Netherlands has to increase to 6.000 MW on land. Construction of additional offshore wind farms are under discussion. The Central Bureau for Statistics reported a total wind electricity production of 3.981 GWh of which 3.308 GWh was produced on land and 679 kWh offshore.

4.17.3 STATUS: SOLAR ENERGY

The contribution of solar energy to the total end use of renewable energy in The Netherlands is small (1.4% of the total end usage of renewable energies). The installation of solar power installations increased last year; in 2010 about 20 MW were installed. The total contribution of solar energy as renewable energy is about 0.3%. Solar energy systems are realized with and without government subsidy. In 2010 a total subsidized budget of 69 MW was available of which 18 MW was installed by March 2011. The subsidy assignments are progressing slowly and chances to get subsidy are low. A project can be planned only after getting subsidy approval: this explains the delay in growth. Furthermore prices of solar energy systems are decreasing, therefore it is attractive for many contractors to wait before starting a project.

4.17.4 GRID MANAGEMENT OUTLOOK

The switch to sustainable energy systems will require radical adaptations to be applied to the Dutch electricity distribution system as many sustainable energy sources do not provide electricity on a continuous basis. Wind and solar energy are dependent on weather conditions and/or the time of the day. Decentralized small-scale electricity generation with solar panels, biogas plants and micro-heat cogeneration units, for example, which are going to replace the central heating boilers, also means that the electricity companies’ hold upon electricity generation is becoming smaller. In addition, the network will be used for ‘two-directional transmission’. As a result of this, the

33 Lead author: Alessandra Liberto

136

energy supply is going to vary widely according to time and place. Nevertheless, the network will have to guarantee a constant voltage and frequency for consumers. One element of the new ‘intelligent’ networks is an effective exchange between the various generation areas, frequently across national borders. This process is supported by the liberalization of the energy market: energy is increasingly transported internationally and as a result the connections between the various countries have been greatly improved. On average, the transport distance for the electricity across the network will continue to increase. This effect would be increased even further if we were to start generating solar electricity on a large scale in sunny areas such as the Sahara. In order to limit the loss of energy in this process, new distribution systems are needed, with a higher voltage, which make use of more conductive materials, or which provide a direct current instead of an alternating current. All of these developments still require much research, amongst others into operating systems and new, safe materials that do not pose any health risks.

Figure 38: Share of renewable energy from national energy sources in the Netherlands (in % of the total electricity consumption in NL). Source :CBS Stateline NL

4.17.5 NATIONAL R&D ACTIVITIES

• ECN, Energy Research Center for The Netherlands (with emphasize on renewable energy) www.ecn.nl • NODE, Nederlands Onderzoeksplatform voor Duurzame Energievoorziening (Dutch research center for renewable energy) www.energieplatform.nl

137

• TNO, TNO: Netherlands Organisation for Applied Scientific Research www.tno.nl

References to §4.17:

[1] Dutch central bureau of statistics - Centraal Bureau voor de Statistiek, NL : Report ”renewable energy in the Netherlands 2010” – www.cbs.nl

138

4.18 NORWAY 34

4.18.1 COUNTRY’S RENEWABLE ENERGY POTENTIAL

The electric power supply in Norway is practically 100 % based on hydro power, a renewable source. Nevertheless, other renewable energy resources are investigated, such as wind, tidal, ocean waves, osmotic power, etc. Only wind power is so far developed up to some significant contribution in the Norwegian electric supply.

4.18.2 STATUS: WIND ENERGY

Wind energy production in Norway continued to grow slightly in 2011 as a result of a previous extension of an investment support program for some wind projects. Further wind power development will be supported by a common green (or renewable) certificate market for Sweden and Norway, beginning January 2012.

Total installed wind power capacity in Norway at the end of 2011 is 540 MW and growth for 2012 is projected to be an additional 10%, at 53 MW. (URL:http://www.vindkraft.no/Default.aspx?ID=148 . (Archived at http://www.webcitation.org/63BgrQ1RO)). The total energy production from wind power in 2010 was 980 GWh. (URL:http://www.vindkraft.no/Default.aspx?AreaID=2 . (Archived at http://www.webcitation.org/63BhTpdJI )). This is slightly under 1% of total Norwegian electricity production of 124.000 GWh. (URL:http://www.ssb.no/emner/01/03/10/energiregn/tab-2011-05-23-07.html . (Archived at http://www.webcitation.org/63BhkMY6o ))

4.18.3 STATUS: SOLAR ENERGY

Solar energy in terms of electric power production was never analyzed in Norway. This is because the energy potential is too low to be of commercial interest. The market for solar panels in Norway is restricted to holiday homes [1].

4.18.4 STATUS: GRID MANAGEMENT

The overall system for Norwegian power supply is based on the N-1 criterion on all levels: if one electric power line is, or has to be taken out of the grid, this shall have no effect on the overall supply. In practice this means in many cases that even two or more components may be taken out of service without noticeable consequences for the system operation. However, there are several bottlenecks in the grid which may be loaded up to its limits in cases with high consumption or high demand for international exchange of electric power. In such cases the transmission lines are operated according to pre-defined current limits (“book rating”) or some are instrumented to monitor the conductor temperature and sag of conductors in some areas. The Norwegian power grid company Statnett SF performed 8-10 years ago a test studies to see if local scale weather forecasts could improve the line ratings (“dynamic rating”). It was however concluded that the accuracy of wind forecasts was not acceptable in the lower range of wind speeds (less than 1 m/s). In addition to this Norway has participated in international collaboration within Cigré (the International Council on Large Electric Systems), especially in the preparation of [2].

4.18.5 NATIONAL R&D ACTIVITIES

34 Lead author: Svein Fikke

139

ICEWIND – Improved forecast of wind, waves and icing. The Norwegian Meteorological Institute and Kjeller Vindteknikk are Norwegian partners in this Nordic project with the overall objective to support the development and integration of wind energy in the five Nordic countries by focusing on three main areas: 1) Icing on wind turbines (atlas, forecasting and losses). 2) Integration of wind energy on land (Iceland). 3) Offshore wind energy (forecasting and access) (http://www.toppforskningsinitiativet.org/en/programmer-1/program-4/prosjekter/icewind )

Icing on overhead power lines is frequently observed in Norway and may cause severe disturbances to the power supply. Integration of wind energy in the system often requires power lines in remote mountainous areas where icing severity needs to be assessed. There are several ongoing research projects (ICEWIND, PhD, etc.) at the Norwegian Meteorological Institute in the field of icing modeling and the use of NWP models for icing prediction.

A list of “Centres for Environment-friendly Energy Research” is provided by The Research Council of Norway on their web page: (http://www.forskningsradet.no/servlet/Satellite?c=Page&cid=1235738786992&pagename=energisenter%2FHoveds idemal ).

References to §4.18:

[1] The Norwegian Research Centre for Solar Cell Technology. Personal communication November 2011. [2] Guide for selection of weather parameters for bare overhead conductor ratings. Cigré TB 299. WGB2.12 August 2006.

140

4.19 POLAND35

4.19.1 THE POTENTIAL OF RENEWABLE ENERGY IN POLAND

Poland is located in Central Europe. The country’s area is 312’679 km2 and the population is about 38 million. Geographical zones in Poland are located in parallel: the northern boundary is mostly along a shore of the Baltic Sea, and the southern part consists of the Sudety and Karpaty mountains. The prevailing part of the country is lowlands.

After Poland has joined the European Union, the Polish legislation on energy was adapted to European regulations. Climate package "320" implied the limitation of emission of carbon dioxide by 20% until 2020, planned reduction of energy consumption by 20% and increase in energy consumption from renewable energy sources by 20%. It is planned to gain an national amount of 15% of energy from renewable sources by 2020.

Table 13: Poland – renewable sources energy – march 2011 [1]

Installation type Quantity Power[MW]

biogas power-stations 149 87.8

biomass power-stations 19 393.0

photovoltaic power-stations 4 0.1

wind power-stations 453 1351.9

hydro-electric power-stations 737 946.3

co-fired technology 42 0.0

TOTAL 1404 2779.1

Traditional sources of energy in Poland are coal and lignite. Their share in electricity production in 2010 was about 90%. It is planned to build a nuclear power plant by 2030. The total amount of installed power capacity in the national grid of electricity in 2010 was 3’575 MW and production of electricity was 157’414 GWh [1]. In connection with EU directives, significant increase in production of electricity from renewable sources in recent years could be seen. The share of electricity generated from renewable sources has increased from 2.6% in 2005 to 7.5% in 2010. Rapid growth occurred since 2005 when favourable regulations for investments in renewable energy sources – so- called "green certificates" – has been introduced. The President of the Energy Regulatory Office issued fifty thousand certificates for the electricity production from renewable energy sources up to October 24th, 2011 [1]. Starting from the year 2005, there was an exceptionally large increase noticed in the installations of wind turbines so

35 Lead author: Katarzyna Starosta

141

that wind energy production capacity has increased from approximately 80 MW in 2005 to about 1500 MW until September 2011. In 2010, the quantity of energy generated by wind farms exceeded the amount of energy produced by regular hydro-power plants. Table 13 shows the current number of installations of renewable energy sources and electric power generated by these objects.

Referring to the Directive of the European Parliament 2009/28/WG of April 23th, 2009, the share of energy from renewable sources specified for Poland was set on a level of 15% of total energy production in 2020. The national action plan for the energy production from renewable sources was accepted by Polish Government on December 7th, 2010 (Table 14). Work on a draft law on renewable energy sources is ongoing.

Table 14: The goals set up in the document "National Plan of Action on Renewable Sources of Energy " [3]

Year 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

% of electricity from RE 7.5 8.8 10.2 11.1 12.2 13.0 13.8 14.7 15.6 16.8

Table 15: Estimation of energy from renewable sources [2]

Description 2011 2015 2019

MW GWh MW GWh MW GWh

Hydropower 962 2 311 1 002 2 439 1042 2567

Solar energy 1 2 2 2 3 3

Wind energy 1550 3255 3 540 7541 5620 12315

Biomass + biogas 450 7110 1530 9893 2230 12943

TOTAL 2 963 12 678 6 074 19 875 8 895 27 828

4.19.2 WIND ENERGY

Wind energy resources for Poland were assessed in "Wind Energy Resources and Zones in Poland" [7]. The most favoured regions regarding wind energy resources is a coast of Baltic Sea and north-eastern part of Poland (the Suwalki region) and the middle and south-eastern part of the country (especially lowlands of Wielkopolska and Mazovia). Wind energy resources on an annual basis in these regions are available at a level at least 1’250kWh/year

142

for 1m2 of a windmill surface [7]. At present time, most of the wind farms are located in the north-western part of the country, few in the middle and in south-eastern regions.

Wind energy is a very dynamically developing branch. The amount of energy produced by wind turbines have increased more than ten times since 2005 until today.

4.19.3 SOLAR ENERGY

The annual volume of incoming radiation for Poland is about 900-1200 kWh/m2 and is similar to many European countries at similar latitudes. The annual course of radiation shows a very non-homogeneous distribution. Average annual sum of hours with sunshine is about 1500 (65 days) and varies spatially. The average daily sunshine duration varies from less than 1 hour in December to about 8 hours from May to July. Values of daily radiation vary in a wide range of less than 1 kWh/m2 (from November to January) to about 10 kWh/m2 from May to August [4].

There is a growing interest for installing solar panels related also to the promotion of environmental trends: this energy is used locally at the level of city quarters, communes and households. Because of relatively easy installation of solar systems, this kind of implementation can become a significant alternative for the production of the missing clean energy.

4.19.4 GRID MANAGEMENT

The energy network in Poland is obsolete and poorly developed. The Energy Law [6] obliges energy companies to purchase energy from renewable sources. Due to the rapid development of renewable energy and the large variability of this kind of energy resource, it is necessary to plan and to predict the energy production. Polish Wind Energy Association has organized on November 4th 2011 a workshop on wind energy in a power network which was attended by specialists from Poland and from abroad. Application of new technical standards can help to reduce the variability in energy usage. The goal is to provide energy to the system which seamlessly works with other sources. Solving this task is an objective for Poland in the near future.

4.19.5 NATIONAL ACTIVITIES

It is mandatory to fulfil the law abiding regulations to support a development of renewable energy production. The efficiency of these new sources depends on highly-variable weather conditions. Purpose of the work of our team of IMGW-PIB as part of the COST Action ES1002 is to provide such quantitative and reliable forecasts.

Concerning the meteorological forecasts, the COSMO and ALADIN models, radar data and measurements from the meteorological network (including solar radiation) stations are the main sources of data for studies. The work should develop new techniques of forecasting and nowcasting dedicated especially for the wind fields. A comprehensive use of results of numerical models, radars, and measurement data via a specialized post-processing, and oriented to customers’ needs should significantly improve the short-term forecasts.

References to §4.19 (all reports in Polish):

[1] Energy Regulatory Office ( http://www.ure.gov.pl ) [2] Setting-out The Objective for Shares of Electricity Produced from Renewable Energy Sources on The Territory of Polish Republic in Electricity Consumption for Domestic Use for The Years 2010-2019, Warsaw, April 12, 2011 (Report of Ministry of Economy, http://www.mg.gov.pl ) [3] National Plan of Action on Renewable Sources of Energy. The Project. ”Warsaw 2010”. (Ministry of Economy, http://bip.mg.gov.pl/ ) [4] Assessment of the resources of solar energy in Poland, Warsaw 2010, IMGW Report

143

[5] The Polish Wind Energy Association http://psew.pl/ (Wind Energy Association) [6] Journal of Regulations of 1997, No 54, pos. 348 [7] Wind Energy Resources and Zones in Poland Halina Lorenc, IMGW-PIB Report

144

4.20 ROMANIA36

4.20.1 COUNTRY’S RENEWABLE ENERGY POTENTIAL

Romania has a population of 22.2 million (2009). The energy generation capacity (2005) is as follows: Nuclear (0.71 GW), Thermal (12.23 GW), Hydro (6.28 GW) and other sources with a total of 19.22 GW (CIA, 2010). The potential of the Romanian renewable energy sources is synthetically presented in Table 16 according to the “National Strategy for Renewable Energy Sources Utilization”.

Table 16: Energy potential of the Romanian renewable energy sources

Renewable energy source Annual energy potential

Solar energy - heat - photovoltaic 60x106 GJ 1,200 GWh

Wind energy 23,000 GWh

Hydro energy, of which, under 10 MW 40,000 GWh 6,000 GWh

Biomass 318x106 GJ

Geothermal energy 7x106 GJ

The average solar radiation in Romania ranges from 1.100 to 1.300 kWh/m2 per year for more than half of the country’s surface. If the solar resource in Romania was used solely for solar thermal applications, the country would have a potential of 60 PJ per year. Romania’s solar electricity potential is approximately 1’200 GWh (Ukraine Biofuel Portal, 2007). The energy share of the solar-thermal systems for the supplying of the necessary amount of heat and hot water for domestic utilization has been estimated to amount to about 1’434x103toe (60PJ/year). These systems could ensure approximately 50% of the hot water production for domestic utilization or 15% of the heat for current heating.

Romania is considered to have the highest wind energy potential in the region, with a predicted total installed capacity of 14’000 MW. Considering the preliminary evaluations in the Black Sea shore area, including the offshore part, the wind energy potential that could be developed on short and medium term may be of about 2,000 MW, with an average electricity production of about 4,500 GWh/year.

There are also good opportunities for biomass development, building off a very large base of existing capacity (over 4,000 MWth). Assuming an available biomass energy supply, district heating systems represent the most immediate and low-cost biomass application – especially CHP plants, industrial co-generation and co-firing.

Romania is rich in water resources. The hydropower potential technically developable of the main rivers amounts to about 36 TWh/year, out of which 23 - 25 TWh/year are economically affordable in great hydropower developments and around 6 TWh/year in small hydropower developments (<10 MW). Romania has a total of at least 767 hydroelectric power plants. A majority of these plants (621) are small hydroelectric plants, with less than 10 MW of capacity.

36 Lead author: Viorel Badescu

145

Romania has the third highest geothermal potential of European nations, with major potential locations on the Western Plain, South Plains in the region of Bucharest, and in the Carpathian regions. The geothermal energy reserve that could be currently tapped in Romania is of about 167x103toe (7,000x106 GJ/year). At present time, Romania has a total geothermal installed capacity of about 145.1 MWt producing 2,841 TJ/year. The energy delivered is of about 30.171x103toe (1,326x106 GJ/year) with an annual average utilization level of 22.3%.

4.20.2 STATUS: WIND ENERGY

Installed wind capacity for the country is approximately 2.5 MW (UDI, 2009). However, Romania currently has approximately 636 MW of wind capacity under construction. The main part of the capacity under construction is located in the southeastern region of Dobrogea, 17 km from the Black Sea, in the Fantanele and Cogeleac wind park with 600 MW capacity. This park will account for approximately 30 percent of Romania’s renewable energy (Realitatea, 2009).

4.20.3 STATUS: SOLAR ENERGY

At the end of 2009, the installed PV systems have had a total power of less than 120 kW. In March 2011, this power raised to 1940 kW, with 1330 kW connected to the grid system (Teodoreanu 2011). The owners are universities, research institutes and local authorities (county councils and city halls). The Romanian Energy Strategy stated as a target for the PV systems 260 MW installed power by 2020 (MECMA 2011). The tendency now is to implement larger grid-connected PV systems (ANRE, 2010).

4.20.4 STATUS: GRID MANAGEMENT

Wind parks with a capacity of 16,000 MW are to be developed in Romania so far. However, only a total capacity of 500- 600 MW has been given technical assent to get connected to the grid, because only 3,000 MW may be injected in the current system. Thus, the requirement for wind energy in Galati (east) amounts to 1,400 MW, while the network’s capacity ranges between 100 and 150 MW only. Under these circumstances, many investors will have to build their own transport lines to have the energy injected in the grid.

4.20.5 NATIONAL R&D ACTIVITIES

Variations of solar irradiance have a significant influence on electric power generation by solar energy systems. The forecast of solar power is a key issue for integration of the solar power production into the electricity grid systems. For this purpose, a proposal to the Romanian Executive Unit for Financing Higher Education, Research, Development and Innovation (UEFISCDI) has recently been submitted and approved to be funded under the project name “Nowcasting and forecasting of PV power plant operation”.

The project aims to implement and develop procedures for nowcasting and forecasting the solar energy input and the PV plant output power in Romania. To this end, two different methodologies will be applied with the final purpose of appropriately integrating the results that will provide better information that those provided by the individual techniques. Particularly it is planned to integrate the results provided by:

• Techniques based on local approaches, such as ARIMA modelling [9], [10], [11]; • Techniques based on Numerical Weather Prediction, in particular Mesoscale Meteorological Models, such as MM5 and WRF.

The objectives of the project are: 1) To evaluate the reliability of local methods, such as ARIMA nowcasting; 2) To evaluate the reliability of the MM5 and WRF to estimate the components of the solar resources in spatial scales of a few kilometres; 3) To provide a deterministic short-term forecasting of the solar radiation and PV output power for the next 48 hours integrating the results from both the MM5/WRF and local estimates;

146

4) To provide, using MM5/WRF, a probabilistic medium-range, 1 to 3 days, forecast of solar radiation and PV output power at daily resolution; 5) To provide, using MM5/WRF, a probabilistic monthly and seasonal forecast of solar radiation and PV output power.

References to §4.20 :

[1] ANRE (2010) Autoritate Nationala de Reglementare in Domeniul Energiei. Raport privind rezultatele monitorizării pieţei de energie electrică în luna decembrie 2010, 3. Structura de producţie a sistemului energetic naţional pe tipuri de resurse, p.5, http://www.anre.ro/ documente.php?id=898; Accessed at: 11 June 2011. [2] CIA (2010) World Factbook, U.S. Energy Information Administration, United Nations Conference on Trade and Development; https://www.cia.gov/library/publications/the-world-factbook/geos/cd.html [3] Energy Information Agency (EIA) (2007) Romania Country Energy Balance; http://www.eia.gov/emeu/world/ country/ countrybal.html [4] MECMA (2011) Ministerul Economiei, Comertului si Mediului de Afaceri. Strategia Energetica a României pentru perioada 2007 – 2020, adoptata prin H.G. Nr. 1069 din 5 septembrie 2007 si publicata in M.O. Nr. 781/19.XI.2007, PARTEA I, Cap.3.7. Prețuri și tarife pentru energie – efecte economice si sociale, Energie electrică, p.14, http://www.minind.ro/anunturi/ strategia_energetica_a_romaniei_ 2007_2020.pdf; Accessed at 11 June 2011. [5] Realitatea (2009) CEZ to cut ribbon in July for first turbine of the 600 MW wind park in Constanta, http:// www.realitatea.net. [6] D. I. Teodoreanu (2011) personal comunication, March 2011 [7] UDI (2009) World Electric Power Plants Database; http://www.platts.com/Products/worldelectric powerplants database [8] Ukraine Biofuel Portal (2007) Renewable Energy Potential in Romania; http://pellets-wood.com/renewable-energy- potential-in-romania-o422.html [9] M. Paulescu, V. Badescu, New approach to measure the stability of the solar radiative regime, Theor Appl Climatol, 2011, 103:459–470, DOI 10.1007/s00704-010-0312-9 [10] V. Badescu, M. Paulescu, Statistical properties of the sunshine number illustrated with measurements from Timisoara (Romania), Atmospheric Research, 101 (1-2), 2011a, 194-204, DOI: 10.1016/j.atmosres.2011.02.009 [11] V. Badescu, M. Paulescu, Autocorrelation properties of the sunshine number and sunshine stability number, Meteorology and Atmospheric Physics, 2011b, 112:139-154; DOI: 10.1007/s00703-011-0135-y

147

4.21 SPAIN37

4.21.1 RENEWABLE ENERGY POTENTIAL IN SPAIN

Spain is one of the leading countries in the world regarding the penetration of renewable energy. Particularly, by 2010, hydro power installed capacity totalized 17 GW, wind energy 20 GW, solar PV 3.7 GW, solar CSP 0.7 GW and biomass about 1 GW [1]. This means about 42 GW for a total installed capacity of the Spanish electric system of 99 GW (about 40%). This renewable energy capacity covered about the 36% of the electricity demand (275 TWh) in 2010 [2]. This number includes 16% for wind energy, 16% for hydro power (highly variable depending on the year), 2% for solar (both PV and Concentration Solar Power) and 2% for biomass.

Figure 39: (Left) Installed capacity by technologies (in %) of the Spanish electric. (Right) Contribution to the electricity demand of different technologies (in %). Source: Spanish Grid operator www.ree.es.

It should be noted that the renewable energy installed capacity has increased considerably in Spain along the last years (Figure 39). As a consequence, the contribution of renewables to the electric demand has also increased considerable in Spain in the last years, from about 19% in 2006 to the current 36%.

Figure 40: Evolution of the electric installed capacity in Spain, for different technologies, along the period2006-2010. Source: Spanish Grid operator www.ree.es.

37 Lead author : David Pozo Vasquez

148

4.21.2 STATUS: WIND ENERGY

Spain is one of the leading countries in the world in wind energy. In comparative terms, Spain is the fourth country in the World (second in Europe) by installed capacity after China (42 GW, 21.8% of the world total), U.S.A. (40 GW) and Germany (27 GW) [3].

Figure 41: Percentage of the world wind energy installed capacity by 2010. Source: Spanish Wind Energy producers association (http://www.aeeolica.org/)

Wind energy was the renewable with the fasted growth in Spain in the last decade (Figure 41). Particularly, in 2010 wind energy production was about 43.700 GWh, contributing with about 16% to the total electric demand on average, and contribution more than 50% during some hours along the years [4]

Figure 42: Evolution of the wind energy production (green) and share (blue) of the wind power in Spain along the last decade. Source: Spanish Wind Energy producers association (http://www.aeeolica.org/).

The wind energy potential in Spain is considerable but at the same time highly dependent on the evolution of the technologies, particularly for the offshore wind energy. Particularly, it is assumed that onshore wind potential in Spain is about 330 GW and the offshore one about 8 GW in shallow water (about 50 m depth as maximum) [5]. In addition, small turbines potential is almost unexplored.

The new National Renewable Energy Plan 2011-2020 [6] aims to increase the onshore installed capacity from the current 21 GW to 35 GW in 2020. At the same time, the plan for first time promotes the offshore wind energy in Spain, aiming the installation of 750 MW by 2020.

4.21.3 STATUS: SOLAR ENERGY

Solar photovoltaic (PV) energy

149

PV solar energy installed capacity in Spain was 3.807 MW by the end of 2010. This power produced 6.279 GWh (about 2% of the total demand). Spain has experienced an enormous growth in solar PV energy in the last decade (Figure 42). Particularly, installed capacity increased from less than 100 MW in 2006 to more than 2.5 GW in 2008 [7]. Then, the change in feed in tariff reduced considerable the growth.

Figure 43: Evolution of the solar PV power installed capacity (yellow) in Spain along the last decade. Source: Spanish PV Energy producers association (http://www.asif.org).

Contribution to the total electricity demand was about 2% in 2010, but this contribution shows a high variability along the year. Particularly, during summer contribution reach up to 4% (Figure 43) [8].

Figure 44: Contribution of the PV power to the electricity demand as a function of the year month, for the year 2009 (red) and 2010 (green). Source: Spanish PV Energy producers association (http://www.asif.org).

The new National Renewable Energy Plan 2011-2020 [9] aims to increase the solar PV installed capacity from the current 3.8 to 7.5 GW by 2020.

Concentrating Solar Power (CSP) energy

Spain is probably the leading country in the world in the development of the Concentration Solar Power (CSP) for electricity production. By 2010, installed capacity totalized 632 MW that produced 691 GWh (0.27% of the total). It should be noted that this technology has only emerged in the last 5 years, being Spain, along U.S.A. the only two countries that has currently commercially operating CSP power plants. Currently, there are plans for developing CSP in many countries of the north of Africa, Italy, Australia, China and India. These plans amount around 2 GW [10]. In Spain, the first commercial plant was operated in 2007, and since then, sustainable growth in the installed capacity has been observed (Figure 44). Particularly, in 2011, new plants totaling 420 MW entered in operation; therefore, by the end of 2011 the installed CSP

150

capacity in Spain was about 1 GW. This installed capacity produced about 1.5% of the total electricity demand by the summer of 2011. By the end of 2013, it is expected that installed capacity reaches 2.5 GW [11].

Figure 45: Evolution of the solar CSP power installed capacity (blue) and their contribution to the electricity demand (yellow) in Spain along the last years. Source: Spanish CSP Energy producers association (http://www.protermosolar.com).

The new National Renewable Energy Plan 2011-2020 [12] aims to increase the solar CSP installed capacity from the current 1 GW to 4.8 GW by 2020.

4.21.4 STATUS: GRID MANAGEMENT

In Spain, the transmission system operator is a private company (under state concession): Red Electrica de España (www.ree.es). The distribution network is operated by several the private utilities: ENDESA (40%), IBERDROLA (40%) and UNION-FENOSA (15%) being the most important. In Spain there is a feed-in tariff system for the remuneration of the production by renewable energies. Independent Power Producers can participate directly in the electricity market. Therefore, wind and solar forecasts are important both for the TSE and the private companies.

One of the main problems of the Spanish electric network is the relatively low level of connectivity, reaching only 6% of the installed power (99 GW). There are certain connection to Portugal and Morocco, but connection to Europe through France has a very low capacity. This is one of the main problems for the development of the renewables in Spain. Many renewables companies are pushing for increase the connectivity to Europe.

4.21.5 NATIONAL R&D ACTIVITIES

Spain has several research organizations and companies that have active initiatives in the field of solar and wind energy forecasting. Spain, particularly, was one of the most promoters of the wind energy forecasting activities, with an important participation in the ANEMOS and SAFEWIND projects. Activities in Solar field are more recent, but many groups are involved currently in solar power forecasting.

References to §4.21:

[1] REE, 2010. Informe anual REE. www.ree.es [2] REE, 2010. Informe anual REE. www.ree.es [3] AEEC, 2011. Estudio Macroeconómico del Impacto del Sector Eólico en España. www.aeeolica.org/ [4] AEEC, 2011. Estudio Macroeconómico del Impacto del Sector Eólico en España. www.aeeolica.org/ [5] IDAE, 2011. Plan Nacional de Energías Renovables (PER) 2011-2020. www.idae.es. [6] IDAE, 2011. Plan Nacional de Energías Renovables (PER) 2011-2020. www.idae.es. [7] ASIF, 2011. Informe anual de la industria fotovoltaica en España. www.asif.org. [8] ASIF, 2011. Informe anual de la industria fotovoltaica en España. www.asif.org. [9] IDAE, 2011. Plan Nacional de Energías Renovables (PER) 2011-2020. www.idae.es.

151

[10] PROTERMOSOLAR, 2011. Impacto macroeconómico del Sector Solar Termoeléctrico en España. www.protermosolar.com [11] PROTERMOSOLAR, 2011. Impacto macroeconómico del Sector Solar Termoeléctrico en España. www.protermosolar.com [12] IDAE, 2011. Plan Nacional de Energías Renovables (PER) 2011-2020. www.idae.es.

152

4.22 SWITZERLAND38

4.22.1 SWITZERLAND'S RENEWABLE ENERGY POTENTIAL

Extending across the north and south side of the Alps, Switzerland encompasses a great diversity of landscapes and climates on a limited area of 41'285 square kilometers. The population is about 7.9 million. The more mountainous southern half of the country is far less populated than the northern half.

In 2010, the total energy consumption of Switzerland was 253 TWh which was the highest value ever. 24% or 60 TWh of this consumption was electrical energy. The total production of electrical energy in 2010 was 66 TWh 38% of which were produced by nuclear power plants. Switzerland has a considerably high portion of renewable energy production due to its hydro power production which accounted for 56.5% of the country’s electricity production in 2010. 0.7% of the total electricity production originated from new renewable energies. 37 GWh were produced by wind energy and 83 GWh by 39 photovoltaic systems . At the end of 2010, the installed capacity was 42 MW for wind energy and 111 MWp for solar energy.

Table 17: Estimated development of renewable electricity production in Switzerland40,3.

Effective Estimated Estimated Estimated Electricity Electricity Electricity Electricity Production Production Production Production [TWh] 2010 [TWh]41 [TWh]2 [TWh]2 2020 2035 2050

Hydropower 37.500 45.340 47.990 47.570

Biomass 1.138 1.642 2.176 3.833

Wind energy 0.037 0.535 2.929 4.000

Solar energy 0.083 0.535 2.929 10.397

Geothermal 0.000 0.276 1.084 4.378

In 2011, the Swiss government decided the nuclear phase-out, i.e. not to build any new nuclear power plants and to successively take the existing nuclear power plants off the grid by 2034. According to the Swiss energy strategy, the missing nuclear energy production will be compensated through four pillars: increased energy efficiency, additional hydro power plants, higher use of new renewable energy resources and possibly new gas power plants (including CHP). The growth of new renewable energies will be pushed by simplified planning procedures and an updated national feed-in law which actually exists since 2007. Table 17 shows the effective electricity production in 2010 and the estimated electricity production until 20502.

38 Lead author: René Cattin 39 Numbers from "Schweizerische Gesamtenergiestatistik 2010" 40 Confirmed by "Energiezukunft Schweiz" study by ETHZ 41 From "Energieszenarien für die Schweiz bis 2050", Zwischenbericht II, 18. Mai 2011, Szenario "neue Energiepolitik", Angebotsvariante 2, Variante C&E

153

The numerous pump hydro plants in the Swiss mountains already act as a "battery for Switzerland" and will presumably play an important role as a “battery for Europe" in the future.

4.22.2 WIND ENERGY

The main wind potential areas in Switzerland are located in the Jura Mountains. In this region, average annual wind speeds up to 7.5 m/s can be found. A good overview can be seen under www.wind-data.ch. Other favorable areas are at the end of the Rhone valley near Martigny and on Alpine passes. However, there are some technical and social obstacles for wind energy development in Switzerland: difficult access, remote grid connections, turbulences and atmospheric icing are technical challenges. Social acceptance, especially regarding visibility and noise exposure are frequently discussed items in the densely populated country.

Wind energy production forecasts are not yet very widely used mainly due to the currently limited installed capacity. However, with increased installed capacity, wind energy production forecasts will become more important especially for actors on the energy spot market. Furthermore, in areas with weak grids (typically in the Jura Mountains), these forecasts might become important in order to maintain the grid stability. An additional challenge for wind power forecasts is the inclusion of atmospheric icing as this can directly affect the energy production. Today, most end users rely on forecasts of wind speed and wind direction and convert these values to energy production internally.

A research project investigated the potential of several downscaling methods for wind power forecasts for complex terrain like in Switzerland. Direct model output of the numerical weather prediction model COSMO was compared to a dynamical downscaling approach, a post-processing approach based on linear regression and a post-processing approach based on a Kalman filter. The results show that the statistical approaches give the most accurate forecasts for most of the locations.

4.22.3 SOLAR ENERGY

Switzerland has a good potential for the production of solar energy. According to the Swiss Federal Office of Energy42, up to 20% of the electricity consumption could be covered by PV electricity production. As there will be many small solar energy systems installed on roofs, the main challenge for the grid is expected to be on the lowest grid voltage levels which are currently not designed for electricity transport in both directions. Solar plants in open areas are currently not in the focus.

Solar energy production forecasts are not yet very widely used mainly due to the currently limited installed capacity. However, they will become more and more important in this country in the near future. Main challenges are the complex terrain as well as snow cover and situations with fog in the low lands. Today, most end users rely on regionally aggregated forecasts of global radiation and convert these values to energy production internally.

4.22.4 GRID MANAGEMENT

Due to the increasing electricity demand, the existing power line capacities in Switzerland approach more and more their limits, while projects for new power lines often face fierce resistance. Therefore, it has become important to exactly know the capacity limits of a given power line under the current meteorological conditions. The maximum current that a cable can carry without damage is assessed in Switzerland by static rating i.e. by ratings for three different maximum air temperatures: +10°C in winter time, +40°C in summer time and +20°C in transition time periods during spring and autumn. The winter limit is exceeded regularly while the summer limit is rather conservative, resulting in unused capacities. In general, there is room for optimization through a more dynamic thermal rating considering the actual weather conditions (temperature, wind, radiation). Today, forecasts of the maximum air temperature for the next five days for different climatic regions in Switzerland are used in the grid management.

42 http://www.bfe.admin.ch/themen/00490/00497/index.html?lang=de

154

Atmospheric icing on power lines is an important issue in the Jura Mountains and on the large transport lines across the Alps. Some of the power lines through the Alps are heated in critical cases through a short circuit procedure. However, this procedure makes it necessary to turn down the power line for a couple of hours which is very relevant for the grid management and the North-South energy transport.

4.22.5 ENERGY EFFICIENCY

Energy efficiency can be improved by developing weather forecasting tuned for the management of energy consumption e.g. for heating, cooling and lightning in buildings, where predictive control approaches are developed that make use of weather forecasts and optimize the allocation of resources in the building control system. Heating in buildings represents about 80% of the Swiss fossil fuel consumption in the household, industrial and service sector. A recent project studying potential gains in introducing weather forecasts in the control of indoor climate in buildings (OptiControl) showed the high importance of the quality of the short term forecast (first few hours and days) and particularly for the incident solar radiation on differently oriented facades.

4.22.6 PLANNED AND PROPOSED NATIONAL R&D

• A national project for assessing the possibilities of dynamic thermal power line rating as well as the influence of atmospheric icing in Switzerland will hopefully start in 2012. At given pilot power lines, meteorological measurements as well as measurements of conductor temperature will be carried out. The main goal is to better understand the correlation between meteorology and conductor performance and to set up short term high-resolution forecasts of conductor temperature and atmospheric icing.

• A performance evaluation for irradiation sensors used in the Solar Energy field will be carried out at the Baseline Surface Radiation Network (BSRN) Payerne with the support of the COST Action ES1002 “WIRE”. The goal is to compare “standard” instruments for measuring (DNI) to high quality radiation monitoring instruments. This will allow to estimate the performance of such instruments and to verify whether they meet the requirements of the solar energy sector.

• A project proposal for very short term solar energy forecasts (+1-6h) based on quasi real-time radiation measurements and satellite data as well as wind fields from a numerical weather model has been proposed to a national utility.

• A proposal for a National Research Program called "MeteoEnergy" plans to cover the meteorological and climatic aspects of energy production, transport and consumption. If accepted, this program will allow for launching a number of research and demonstration projects in this field.

155

4.23 TURKEY43

4.23.1 TURKEY’S RENEWABLE ENERGY POTENTIAL

Turkey is located in the Southeastern Europe and Eastern Mediterranean region with a population of about 73 million. Turkey has experienced rapid economic growth since 1980s and today ranks among the fastest growing energy markets in the world. The total national installed capacity is about 45’000 MW by the end of 2009, while it was 16’318 MW in 1990 with the average annual growth rate of 5%. Total power generation of the country rose from 57’543 GWh in 1990 to

194’813 GWh in 2009, with an average growth rate of 6.7% [1]. Hence, energy-related CO2 emissions have more than doubled since 1990. Turkey having ratified Kyoto Protocol in 2009, there are several efforts to control and mitigate GHG emissions in the country. In this regard, the utilization of renewable energy resources such as solar, geothermal and wind energy appears to be one of the most efficient and effective ways to achieve the Kyoto Protocol’s requirements.

Turkey does not have enough primary energy sources such as petroleum and natural gas. The second largest energy source after coal and natural gas is hydropower. The Turkish government hopes that hydropower capacity will expand to 35’000 MW by the year 2020.

Table 18: Turkey’s renewable energy potential

Energy Type Usage Purpose Natural Capacity Technical Economical

Electric (billion kWh) 977,000 6105 305 Solar energy Thermal (Mtoe) 80,000 500 25

Hydropower Electric (billion kWh) 430 215 124.5

Wind energy (land) Electric (billion kWh) 400 110 50

Wind energy (off shore) Electric (billion kWh) - 180 -

Wave energy Electric (billion kWh) 150 18 -

Electric (109 kWh) - - 14 Geothermal energy Thermal (Mtoe) 31,500 7500 2.843

Biomass energy Total (Mtoe) 120 50 32

Geothermal energy is the only renewable energy source not originating from solar energy which includes direct use of heat, electricity production and geothermal heat pump. Turkey is the seventh richest country in the world in geothermal potential

43 Lead author: Selahattin Incecik

156

for its direct use and for electricity generation. It is located on the Alpine–Himalayan organic belt, having one-eighth of the world’s geothermal potential [2]. It has 170 geothermal fields over 400 °C which are located in the Western and Central parts of the country [3].

Wind and solar resources represent a massive energy potential, which greatly exceeds that of fossil fuel resources. The usage of these renewable energy resources is a promising prospect for the future as an alternative to conventional energy.

Turkey’s energy policy target is planned to reach a 20% renewable energy share by the year of 2020, and 30% in 2050, respectively [4]. The national Ministry of Energy and Natural Resources of has listed CSP (concentrating solar power) as an important research issue. On the other hand, up to now, no commercial solar thermal power plant is in operation. Turkey’s renewable energy potential is given in Table 18.

Renewable energy production makes up approximately 14.4% of the total primary energy supply (TPES), i.e. 10.30 Mtoe in 2007, and renewable sources represent the second-largest domestic energy source after coal. In spite of this high potential, solar energy is not yet widely spread, except for flat plate solar collectors which are mainly used for domestic hot water production, mostly in the sunny coastal regions. In 2007, about 8.0 million m2 solar collectors were produced and it is predicted that total solar energy production will reach about 0.390 Mtoe [5].

Since mid-2000’s, as a result of applied important structural changes and increased investment incentives in renewable energy, investment demands above the potential on wind energy have increased in Turkey. In the near future there will be an overloaded demand to solar energy like wind energy. In this context the possible usage areas and rate effects of renewable energies replaced with fossil fuels to the economic benefit of the country can be examined.

4.23.2 WIND ENERGY

In Turkey’s history, wind energy has always played an important role in Anatolia. The people used wind energy for hundreds of years for pumping water and grinding grain. Today, it is estimated that Turkey has 160 TWh/year of wind potential, which is about twice as much as the current electricity consumption.

According to Ministry of Energy and Natural Resources, the wind potential reaches 88’000 GW and the economic potential is estimated at 10’000 MW or higher [6] which represents a significant contribution to the energy system. But for effective energy planning in the grid systems, prediction of wind energy production should be obtained accurately up to 48-72 hours.

Today, the majority (25) of wind energy projects in Turkey are concentrated in the west and Mediterranean regions. The installed capacity is expected to reach 20’000 MW by 2023.

4.23.3 SOLAR ENERGY

In recent years, solar energy utilization in various applications has increased significantly. Turkey’s geographical location is highly favorable for the utilization of solar energy.

The country has a daily average of 3.6 kWh/m2 irradiance and 7.2 hours sunshine duration. In the southern parts of Turkey, there are approx. 3’000 sunshine hours per year. The yearly total solar irradiance is 1’311 kWh/m², and the average total yearly sunshine duration is 2640 h. In Southeastern Anatolia and the Mediterranean regions, higher yearly averages closing to 1’500 kWh/m² can be expected.

The main solar energy utilizations in Turkey are the flat-plate collectors for domestic hot water systems. The hot water heating system installations exceeds an area of 10 million m2 with a total installed capacity of 7.8 GWth in 2008. Turkey occupies the 2nd rank of the top countries using solar thermal power worldwide, following China. Utilization of photovoltaic systems remains limited to governmental organizations in remote service areas such as telecom stations, forest fire observation towers and highway emergency only. However, it is expected that the photovoltaic system will play an important role in the future energy planning of Turkey.

157

4.23.4 GRID MANAGEMENT

Development and operation of smart grids for power distribution in Turkey has not been forced yet due to the present inadequate infrastructure. Recently, in wind energy applications, network connection problems and forecasting of electricity is on discussing. For this purpose, a law has been forced in December 2010. Finally, the integration of distributed power generation and renewable energy sources into existing and future unified electricity systems will represent an enormous technological challenge with their flexible structures

4.23.5 A NATIONAL PROPOSAL: SHORT TERM FORECASTING OF SOLAR RADIATION USING WRF MODEL

Variations of solar irradiance have a significant influence on electric power generation by solar energy systems. Forecast of solar power is a key point for integration of the solar power production into the electricity grid systems. For this purpose, a proposal to the Turkish Scientific and Technical Research Council has recently been submitted. In this project it is aimed to forecast the solar radiation for a short time over Turkey by using previously measured solar radiation data and the WRF model used by academicals institutes and weather services. The suggested study entitled “Short term forecasting of solar radiation using WRF model” is planned to be carried on under the COST Action ES1002 “WIRE” label. As a non- hydrostatic atmospheric model, WRF allows the users to bring the regional resolution to a few kilometers: a spatial resolution of 5x5km will be applied. Global radiation, cloud cover, temperature and wind speeds will be used for the months January, March, July and September. The results obtained from the model will be compared with ground based data of the same period and evaluated considering cloud effects. A benchmarking process will evaluate the performance of the developed models. Model Output Statistics (MOS) will be applied to the output of WRF Model to improve the forecast.

References to §4.23:

[1] B. Ozer, E.Gorgun, S. Incecik. An analysis of the CO2 emission mitigation potential from electricity production industry in Turkey. SET2011, 10th International Conference on Sustainable Energy Technologies, Istanbul Turkey, 2011 [2] N. Cicek, M. Ozturk, N. Ozek, Renewable energy market conditions and barriers in Turkey, Renewable and Sustainable Energy Reviews, 2009, Volume 13; 1428-1436 [3] EIE, Electrical Power Resources Survey and Development Administration. Potential of Turkish wind and solar power. Available from http://www.eie.gov.tr [accessed 20.06.09]. [4] MENR, Ministry of Energy and Natural Resources. Energy report of Turkey in 2007. Turkey, Ankara; available from http://www.enerji.gov.tr [accessed 10.06.09]. . [5] Toklu, E., M.S. Guney, M. Isık, O. Comaklı, K. Kaygusuz. Energy production, consumption, policies [6] MENR, Ministry of Energy and Natural Resources. Energy report of Turkey in 2008. Available from http://www.enerji.gov.tr [accessed 04.06.10].

158

5 CONCLUSIONS / RECOMMENDATIONS

During the second half of the 20th century, the problematic of climate change became more and more obvious. Domains such as the effects of the alarming increase of global temperature due to man-made production of greenhouse gases, the decrease of biodiversity, the consequences of frightening pollution levels in cities and many other aspects became more and more the subject of discussions in relationship with political visions and decisions. Central to these discussions was – and still is – the efficient production and distribution of energy and in particular the use of clean renewable energies which will not affect the state of the environment. The Chernobyl and Fukushima nuclear accidents gave new impulses to develop such kind of energy supply sources which will not threaten the future societies.

Before the introduction of renewable energies, the situation was relatively straightforward: the demand on the users’ side could be match by the adequate tuning of the different types of fossil, hydroelectric or nuclear power plants and the electrical grids were adapted to this configuration with centralized energy sources on one side and consumption centers on the other side.

With the increasing penetration of renewable energy sources, and especially for those dependent on the variable weather conditions such as wind and solar energies, the situation is dramatically changing as a third partner is joining the game with intermittent production. Furthermore, these new energies are usually located far away from consumption centers, in often harsh environments. This means that the electrical grid operators will have to cope with variable energy input which cannot be manually switched on and off and which are more geographically distributed that the conventional power sources. In that sense, a wider – European - integration of the energy networks is unavoidable as the wind potential lies mainly in the northern countries while the solar production is located more in the southern regions. Furthermore, the use of new, well adapted storage facilities and of modern smart grid technologies on the very short, short and long-term scales is becoming mandatory to maintain the grid stability.

Introducing intermittent, weather dependent energy sources means also that the power plant operators and electrical grid managers will have to increase their knowledge of the production to be expected in the next future and, in case of geographical integration, from where this energy will be coming. It is therefore mandatory to develop accurate production forecasting techniques able to compute how much and from where the energy will be available in the next future. In other words, it is becoming obvious that the penetration of renewable energies will be linked to the availability of high-level production forecasting tools, to the adaptation of the design and operation of electrical grids together with the development of adequate storage capacities.

Recommendations

The COST Action ES1002 will continue its activities following the goals defined in the Memorandum of Understanding (www.wire1002.ch) :

• Promote the development of forecasting systems for the production of renewable energies. • Support the development of standard and remote sensing measurement to increase the quality of the forecast systems (in-situ fast measurements and data assimilation for very short term forecasting) and to validate the performances of the forecasting systems • Evaluate the performances of measurement systems • Increase the collaboration with the power grid managers to promote the use of accurate forecasting systems for the production of renewable energies

Halfway through the present Action, the following recommendation for future activities can be already expressed:

The increased integration of renewable energies at the European level will be best sponsored by the establishment of a European institution dedicated to renewable energy production forecasts through a specific joint initiative within the forthcoming European programs including:

• The support of R&D activities in this field.

159

• The promotion of the development of operational forecast (NWP) of related variables specific for renewable energies. • The promotion of the development of energy meteorology for specific harsh environmental areas such as mountainous areas and coastal zones. • The promotion of the use and integration in numerical models of modern measurement technologies such as ground based sky imagers, radars, lidars, ceilometers, satellite estimates, etc... • The contribution to the optimal design and control of European electrical grids. • The support of the penetration of renewable energies in technically less developed countries.

160

6 APPENDIX A44

PERFORMANCE EVALUATION FOR IRRADIATION SENSORS USED IN THE SOLAR ENERGY FIELD: DIRECT NORMAL SOLAR IRRADIANCE INTER-COMPARISON AT THE METEOSWISS BSRN PAYERNE STATION

Background

COST ES1002 WIRE is a European action aimed at enhancing meteorological forecasting for renewable energy production. Switzerland plays a leading role in COST WIRE and MeteoSwiss is a partner. Following a specific request from COST WIRE, MeteoSwiss took the responsibility of conducting an inter-comparison of radiometers measuring Direct Normal solar Irradiance (DNI) as contribution to the collaboration (see Trip report COST ES1002 “WIRE” State of the Art Workshop). Different geometries are used for the solar energy collection devices such as solar concentrators, photovoltaic panels with various orientations or solar thermic panels. Thus, the information about the overall solar energy flux over an horizontal surface (global irradiance) is not sufficient for assessing the solar energy input onto the collection device. Instruments that allow inferring separately the solar direct and diffuse radiation components are more adequate since one can reconstruct the radiance distribution with this information and limited assumptions on the distribution of the diffuse radiance. In addition, a measurement frequency on the order of min-1 allows acquiring information on ramps. On the other hand, accuracy requirements are not as strict as for climate monitoring (see “Requirements for radiometers” section below).

Project goals

• Comparing “standard” instruments for measuring (DNI) to high quality radiation monitoring instruments (reference) from the Baseline Surface Radiation Network (BSRN) Payerne site. This will allow estimating the performance of such instruments and verifying that they answer the requisites of the solar energy sector. • Instruments that will be tested (later referred to as target instruments, see below) are instruments that can be operated without high maintenance and whose cost allows the deployment in networks. • The inter-comparison will be limited to the assessment of the instrument performances. However, it represents a first step: it establishes a basis on which subsequent steps will allow defining Standard Operating Procedure (SOP) for the use of such instruments and verifying the adequacy of their performances to the requirements of the solar energy sector. Such subsequent tasks should be carried out in collaboration with instrument users in the solar energy sector (including private partner). COST ES1002 WIRE is the natural framework for such collaboration. Such collaboration will guarantee that the performance evaluation produces results adequate for these subsequent steps.

Deliverables

Sets of instruments from different manufacturers will be compared to the BSRN reference measurements during an intensive operation period (IOP). This will be followed by a long-term performance evaluation period with a subset of instruments tested. After the IOP a preliminary analysis will be conducted and described in a first report. Then, a final report will describe the evaluation results after the long-term evaluation period. For each type of instrument tested, a datasheet summarizing the performances of the instrument will be issued. The instrument performance will be evaluated with respect to the reference for: a) direct normal irradiance, b) diffuse irradiance and c) global irradiance. The performance evaluation will estimate bias and root mean square error (RMSE) for assessing the effective error between the instrument and reference measurements. Effective error45 is here understood as

44 Lead author: Laurent Vuilleumier

45 According to the Guide to the expression of uncertainty in measurements (GUM), the error is the difference between the measurements and the true value of the measurand, which is unknowable. This “effective error” definition is thus not strictly in agreement with GUM error definition, but is used to designate the difference to reference that should be determined by the inter-comparison.

161

estimates of the difference between the measurements from the target instruments and a reference derived from the measurements by the BSRN reference instruments (see Appendix D). This includes a systematic (bias) and statistical part (RMSE). Such error estimates are themselves subject to an uncertainty that will need to be evaluated (see Appendix E). The RMSE dependence on time step will be investigated (the reference shortest time step is 1min, longer integration time will also be investigated). Similarly, the dependence of bias and RMSE on solar zenith angle will be studied for evaluating the quality of the correspondence between the diurnal cycles as measured by the tested instruments and the references. In addition, the dataset will be separated in data recorded during clear-sky, broken cloud coverage and overcast times, and with respect to seasonal cycle. The analysis will be conducted separately on the 3 datasets. Measurement frequency and data availability will also be part of the evaluation criteria. Finally, in collaboration with the Deutsches Zentrum für Luft- und Raumfahrt (DLR), the possibility of testing how well calibration results obtained at one location (e.g., Plataforma Solar de Almería) compare to results at another location (Payerne) will be explored.

Requirements for radiometers used in the solar energy sector

Preliminary enquiries indicate that desirable characteristics or performance from instruments measuring solar radiation for the solar energy sector include: • ability to distinguish the direct and diffuse component of solar irradiance; • time resolution < 15min (1min if studying ramp); • RMSE ≤ 5% (at 10min resolution, no indication yet on tolerable bias); • data availability > 93%.

Target instruments

The tested instruments will be instruments allowing inferring the diffuse and direct component of solar (shortwave) radiation separately in a cost effective way and without costly maintenance. In particular, such instruments do not necessitate the use of sun trackers. Two kinds of such instruments are on the market. The first type uses a rotating shadow-band that alternately shades and then exposes the entrance aperture of the instrument. Such measurement cycles allow estimating the global solar irradiance and the diffuse irradiance component with proper algorithms. The direct-normal component is then computed from the difference of the two measured components. The other type of instrument uses an elaborate computer-generated shading pattern and an array of thermopile sensors. Such instrument is designed in such a way that for almost any position of the sun in the sky, some sensors are exposed to the direct sun and some are in the shade. This also allows inferring the global solar irradiance and its diffuse component, and subsequently the direct normal irradiance. A list of such instruments and their manufacturers is given in Appendix A. Note that different combinations of sensors, correction algorithms and calibrations are used. The complete combination should be considered as the “instrument”. Thus there are potentially more “instruments” than the cited hardware models.

162

7 APPENDIX B46

QUESTIONNAIRE ON REQUIRED PERFORMANCES OF SOLAR IRRADIANCE MEASURING INSTRUMENTS FOR THE SOLAR ENERGY SECTOR.

An inter-comparison is planned in 2012 at MeteoSwiss Aerological Station in Payerne, Switzerland, to evaluate the performance of instruments able to infer the diffuse and direct component of solar (shortwave) radiation separately in a robust and cost effective way and without intensive maintenance, typically for use in the solar energy sector. The target instruments are not maintenance-intensive and do not necessitate the use of sun trackers. They can be used for irradiance monitoring with separation of direct and diffuse components. This allows estimating the radiance distribution for determining the solar energy input on any kind of surface. The inter-comparison will compare measurements by the tested instruments to measurements from collocated high quality radiation monitoring instruments (reference) from the Baseline Surface Radiation Network Payerne site. The reference data are 1min mean, standard deviation, minimum and maximum of 1Hz measurements. Comparisons can be made for any time intervals larger or equal to 1min. Currently the following deliverables are anticipated (very summarized): Instrument performance will be evaluated with respect to the reference for: a) direct normal irradiance, b) diffuse irradiance and c) global irradiance. The error between the instrument and reference measurements will be estimated using bias (systematic) and root mean square error (statistical). This will include an uncertainty analysis on the error. The dependence of bias and RMSE with respect to solar zenith angle will be studied for evaluating the quality of the correspondence between the diurnal cycles as measured by the tested instruments and the references. In addition, the dataset and analysis will be separated in data recorded during clear-sky, broken cloud coverage and overcast times, and with respect to seasonal cycle. In addition measurement frequency and data availability will be evaluated.

46 Lead author : Laurent Vuilleumier

163

8 LIST OF AUTHORS

Agustsson Halfdan, Institute for Meteorological Research, Iceland, [email protected] Badescu Viorel, Romania, [email protected] Batchvarova Ekaterina, NIMH, Bulgaria, [email protected] Blanc Philippe, Mines-ParisTech / ARMINES, France, [email protected] Bonelli Paolo, R.S.E., Italy, [email protected] Calpini Bertrand, MeteoSwiss, Switzerland, [email protected] Cattin René, Meteotest, Switzerland, [email protected] Erpicum Michel, Université de Liège, Belgium, [email protected] Fikke Svein, Meteorological consultant, [email protected] Giebel Gregor, DTU, Denmark, [email protected] Gryning Sven-Erik, DTU, Denmark, [email protected] Heimo Alain, Meteotest, Switzerland, [email protected] Heinemann Detlev, Oldenburg University, Germany, [email protected] Horvath Kristian, Meteorological and Hydrological Service, Croatia, [email protected] Hosek Jiri, Academy of Sciences, Czech Republic, [email protected] Incecik Selahattin, Turkey, [email protected] Kann Alexander, ZAMG, Austria, [email protected] Kariniotakis Georges, MINES-ParisTech / ARMINES, France, [email protected] Kazantzidis Andreas, University of Patras, Greece, [email protected] Kishcha Pavel, Tel Aviv University, Israel, [email protected] Liberto Alessandra, Telvent, Netherlands, [email protected] Niemela Sami, FMI, Finland, [email protected] Pinson Pierre, DTU, Denmark, [email protected] Pozo Vasquez David, University of Jaén, Spain, [email protected] Schroedter-Homscheidt Marion, DLR, Germany, [email protected] Sempreviva Anna Maria, CNR-ISAC, Italy, [email protected] Starosta Katarzyna, Institute of Meteorology and Water Management, Poland, [email protected] Todorovic Jovan, Elektroprenos BiH, Power Transmission Company B&H, Bosnia & Herzegovina, [email protected] Thomaidis Nikos, University of the Aegean, Greece, [email protected] Toth Katalin, Hungarian Meteorological Service, Hungary, [email protected] Vuilleumier Laurent, MeteoSwiss, Switzerland, [email protected]

164