POTENTIAL OF SOLAR PHOTOVOLTAIC AND WIND POWER PLANTS IN

MEETING ELECTRICITY DEMAND IN AFGHANISTAN

Thesis

Submitted to

The School of Engineering of the

UNIVERSITY OF DAYTON

In Partial Fulfillment of the Requirements for

The Degree of

Master of Science in Renewable and Clean Energy

By

Ahmad Murtaza Ershad

Dayton, Ohio

May, 2014

POTENTIAL OF SOLAR PHOTOVOLTAIC AND WIND POWER PLANTS IN

MEETING ELECTRICITY DEMAND IN AFGHANISTAN

Name: Ershad, Ahmad Murtaza

APPROVED BY:

______

Robert J. Brecha, Ph.D. J. Kelly Kissock, Ph.D. Advisory Committee Chairman Committee Member Professor Chair and Professor Mechanical and Aerospace Mechanical and Aerospace Engineering Engineering

______

Kevin P. Hallinan, Ph.D. Committee Member Professor Mechanical and Aerospace Engineering

______

John G. Weber, Ph.D. Tony E. Saliba, Ph.D. Associate Dean Dean, School of Engineering School of Engineering & Wilke Distinguished Professor

© Copyright by

Ahmad Murtaza Ershad

All rights reserved

2014

ABSTRACT

POTENTIAL OF SOLAR PHOTOVOLTAIC AND WIND POWER PLANTS IN

MEETING ELECTRICITY DEMAND IN AFGHANISTAN

Name: Ershad, Ahmad Murtaza University of Dayton

Advisor: Dr. Robert J. Brecha

This thesis analyses the potential of large-scale grid-connected solar photovoltaic

(PV) and wind power plants in two of Afghanistan’s most populous provinces (Balkh and

Herat) to meet a fraction of growing electricity demand. The analysis is performed by quantifying resource quality, variability and cost of energy generation. First, the quality of solar and wind resources is quantified by characterizing wind speed and solar radiation and calculating capacity factors and energy yields from hypothetical power plants using measured wind speed and typical solar radiation data. Second, variability of wind and solar resources is quantified by comparing their daily and seasonal profiles with electricity demand profiles, analyzing their impacts on load duration curves and determining their penetration and curtailment levels for various demand scenarios.

Finally, cost of energy generated from solar PV and wind power plants is determined.

The research shows that future solar PV and wind power plants in Balkh and Herat provinces could achieve very high penetration levels without significant curtailment

iv meaning less reliance on unpredictable and unstable power purchase agreements with neighboring countries, longer life of limited domestic fossil fuel resources such as coal and natural gas, and less imports of diesel fuel with rising costs and unfriendly environmental impacts.

v

Dedicated to Afghanistan.

vi

ACKNOWLEDGMENTS

I would like to thank Dr. Robert J. Brecha, my thesis committee chair and advisor, for his time and guidance for completing this work. In the meantime, I would also like to thank Dr. Kevin Hallinan and Dr. Kelly Kissock, my committee members, for their support, ideas and comments. Not only did they support me academically, they helped me appreciate the importance of energy in development and applauded me again and again for my research about Afghanistan.

It is not easy to do research about Afghanistan without getting frustrated due to lack of adequate and reliable data. However, I was lucky enough to receive some help from colleagues in Afghanistan. I would like to thank Engineer Haris Haidari of

Sustainable Energy Services Afghanistan (SESA) for providing me with the wind speed data used in this thesis.

In the end, I would like to appreciate the support of my parents, my fiancé and the rest of my family and friends. I am very happy to have undertaken this research and may the information presented here be a step toward building an energy independent

Afghanistan.

"And say: My Lord increase me in knowledge." Quran 20:114

vii

TABLE OF CONTENTS

ABSTRACT ...... iv ACKNOWLEDGMENTS ...... vii TABLE OF CONTENTS ...... viii LIST OF FIGURES ...... x LIST OF TABLES ...... xiii LIST OF ABBREVIATIONS AND NOTATIONS ...... xv 1. INTRODUCTION ...... 1 1.1. Background ...... 1 1.2. Literature Review ...... 5 1.3. Research Objective ...... 9 2. RESOURCE QUALITY...... 11 2.1. Datasets ...... 11 2.2. Wind Resource Characteristics ...... 14 2.2.1. Wind Speed ...... 14 2.2.2. Turbulence Intensity (TI) ...... 22 2.2.3. Wind Direction...... 25 2.2.4. Wind Shear...... 26 2.2.5. Wind Speed Distribution...... 28 2.2.6. Wind Power Density ...... 31 2.3. Wind Plant Output ...... 32 2.4. Solar Resource Characteristics ...... 39 2.5. Solar PV Plant Output ...... 42 3. ELECTRICITY DEMAND ...... 47 3.1. Seasonal Profile ...... 48 3.2. Diurnal Profile ...... 50

viii

4. RESOURCE VARIABILITY...... 52 4.1. Seasonal and Daily Correlation ...... 52 4.2. Residual Load Duration Curves ...... 55 4.3. Penetration and Curtailment Levels ...... 58 5. COST OF ENERGY GENERATION ...... 64 6. CONCLUSION ...... 68 REFERENCES ...... 71

ix

LIST OF FIGURES

Figure 1. Map of Afghanistan ...... 3

Figure 2. Share of energy generation sources in 2011 ...... 5

Figure 3. Seasonal variation of wind speed and standard deviation in Uljato at 50 m height...... 15

Figure 4. Seasonal variation of wind speed and standard deviation in Hotel Safid at 50 m height...... 15

Figure 5. Monthly wind speed in Uljato and OAMS ...... 16

Figure 6. Monthly wind speed in Hotel Safid and OAHR ...... 16

Figure 7. Changes in monthly average wind speeds in OAMS for the period from 2011 to

2013...... 17

Figure 8. Hotel Safid and Uljato annual average daily wind speed profile ...... 18

Figure 9. Daily profile of measured wind speed in Uljato and OAMS ...... 19

Figure 10. Daily profile of measured wind speed in Hotel Safid and OAHR ...... 20

Figure 11. Seasonal variation of wind speed diurnal profile in Hotel Safid ...... 20

Figure 12. Seasonal variation of wind speed diurnal profile in Uljato ...... 21

Figure 13. Autocorrelation function of measured wind speed in Hotel Safid and Uljato 22

Figure 14. Seasonal variation of TI in Hotel Safid and Uljato ...... 24

Figure 15. Daily variation of TI in Hotel Safid and Uljato ...... 24

x

Figure 16. Wind speed direction as percent of total energy in Uljato (left) ...... 25

Figure 17. Wind speed direction as percent of total time in Uljato (right) ...... 25

Figure 18. Wind direction as percent of total energy in Hotel Safid (left) ...... 26

Figure 19. Wind direction as percent of time in Hotel Safi (right) ...... 26

Figure 20. Monthly variation of average wind shear coefficient ...... 27

Figure 21. Daily variation of average wind shear coefficient ...... 28

Figure 22. Uljato wind speed frequency distribution at 50 m...... 28

Figure 23. Hotel Safid wind speed frequency distribution at 50 m ...... 29

Figure 24. Actual and bi-Weibull wind speed distribution frequency ...... 31

Figure 25. 630 KW wind turbine composite power curve ...... 33

Figure 26. Actual and fitted power curve of the 630KW wind turbine ...... 36

Figure 27. Monthly net capacity factors of Hotel Safid and Uljato win power plants ..... 38

Figure 28. Annual average daily net energy yield of Hotel Safid and Uljato wind power plants ...... 38

Figure 29. Monthly global horizontal irradiance (GHI) in Afghanistan ...... 39

Figure 30. Map of annual global horizontal irradiance (GHI) for Afghanistan ...... 40

Figure 31. Annual average monthly GHI at OAMS by various models ...... 41

Figure 32. Annual average monthly GHI at OAHR by various models ...... 41

Figure 33. Annual average daily global horizontal irradiance at OAMS and OAHR ...... 42

Figure 34. Monthly net capacity factor for solar PV power plants in Balkh and Herat ... 45

Figure 35. Annual average diurnal AC power output of solar PV power plants in Balkh and Herat ...... 46

Figure 36. Load histogram for 2010 and 2011 ...... 49

xi

Figure 37. Kabul province average monthly demand in 2010 and 2011 ...... 50

Figure 38. Kabul 2010 - 2011 annual average daily demand profile ...... 51

Figure 39. Seasonal variation of intermittent renewables and electricity demand in Balkh

...... 54

Figure 40. Seasonal variation of intermittent renewables and electricity demand in Herat

...... 54

Figure 41. Annual average daily profile of demand versus wind and solar in Herat ...... 55

Figure 42. Annual average daily profile of demand versus wind and solar in Balkh ...... 55

Figure 43. Residual load duration curves (RLDC) of Herat grid after introducing solar PV and wind power independently ...... 57

Figure 44. Residual load duration curve (RLDC) of Balkh grid after introducing solar PV and wind power independently ...... 57

Figure 45. Residual load duration curves (RLDC) of Balkh and Herat grids after introducing solar PV and wind power together ...... 58

Figure 46. Balkh wind power penetration...... 60

Figure 47. Balkh wind power curtailment losses ...... 60

Figure 48. Balkh solar PV power penetration ...... 61

Figure 49. Balkh solar PV power curtailment ...... 61

Figure 50. Herat wind power penetration ...... 62

Figure 51. Herat wind power curtailment losses ...... 62

Figure 52. Herat solar PV power penetration ...... 63

Figure 53. Herat solar PV power curtailment losses ...... 63

xii

LIST OF TABLES

Table 1. ACEP wind monitoring sites in 2012 ...... 12

Table 2. Range test criteria ...... 12

Table 3. Dataset properties for Uljato and Hotel Safid sites ...... 13

Table 4. Operating ranges and accuracy of measurement sensors used by ACEP ...... 13

Table 5. Uljato and Hotel Safid annual average measured wind speeds in 2012 ...... 14

Table 6. IEC 61400-1 Turbulence categories 3rd edition ...... 23

Table 7. Average wind shear coefficients in Uljato and Hotel Safid ...... 27

Table 8. Weibull parameters for measured wind speed at 50 m height ...... 30

Table 9. Bi-Weibull parameters for wind speeds at 50 m in Uljato and Hotel Safid ...... 31

Table 10. Hotel Safid and Uljato wind power densities at 30 m, 40 m and 50 m height . 32

Table 11. Calculated and assumed wind farm losses for Hotel Safid and Uljato ...... 37

Table 12. Net annual energy production and capacity factors for Hotel Safid and Uljato wind power plants and typical capacity factor around the world ...... 38

Table 13. OAMS annual and daily average GHI ...... 40

Table 14. OAHR annual and daily average GHI ...... 40

Table 15. Net annual energy production and capacity factors for Hotel Safid and Uljato solar PV power plants and typical capacity factor around the world ...... 45

Table 16. Kabul Province demand statistics and load factor in 2010 and 2011 ...... 49

xiii

Table 17. Economic parameters used in calculation of electricity generation cost of solar

PV and wind power plants in Balkh and Herat provinces ...... 65

Table 18. Net annual energy production and capacity factors of solar PV and wind power plants in Balkh and Herat ...... 65

Table 19. Electricity generation cost of solar PV and wind power plants in Balkh and

Herat ...... 67

xiv

LIST OF ABBREVIATIONS AND NOTATIONS

ACEP Afghanistan Clean Energy Program ADP Asian Development Bank AEIC Afghanistan Energy Information Center AEP Annual Energy Production AFG Afghani AVG Average AWEA American Wind Energy Association CF Capacity Factor GDP Gross Domestic Product GHI Global Horizontal Irradiance GIS Geographic Information System GW Gigawatts IEC International Electro-mechanical Commission KWH Kilowatt Hours LCOE Levelised Cost of Energy MW Megawatts NEPS North East Power System NRECA National Rural Electric Cooperative Association NREL National Renewable Energy Laboratory OAHR Mazar-e-Sharif International Airport OAMS Herat International Airport PSMP Power Sector Master Plan PV Photovoltaic PVF Present Value Factor RLDC Residual Load Duration Curve TI Turbulence Intensity TMY Typical Meteorological Year United States Agency for International USAID Development USD US Dollars WB World Bank

xv

1. INTRODUCTION

1.1. Background

About 1.3 billion people do not have access to electricity in the world and four out of five of them live in Afghanistan and other developing countries of South Asia and

Sub-Saharan-Africa (REN21, 2013). Worldwide greenhouse gas emissions increased by

70 % between 1970 and 2004 (IPCC, 2007). Fossil fuels are becoming more scarce and expensive. Electricity demand in Asia and Pacific is projected to more than double between 2010 and 2035 (ADB, 2013). In the recent years, countries around the world have felt the need to diversify their power production fuel inputs and adopted renewable energies in response to challenges mentioned above.

At the same time, rapid growth has been seen in the use of renewable energy sources for electricity generation across the world. Annual average growth rate of cumulative installed capacity of solar PV and wind from the end of 2002 through 2012 was 47 % and 25 % respectively (REN21, 2013). Total global renewable power

(excluding hydro) capacity exceeded 480 GW and supplied about 5.2 % of global electricity in 2012 (REN21, 2013). Total global installed capacity of solar PV and wind power reached 100 GW and 283 GW respectively by the end of 2012 (REN21, 2013).

China, the United States, Germany, Italy, Spain and India are the top countries for renewable energy power capacity and renewables represented about 11 % of total

1 installed capacity in India by 2012 (REN21, 2013). Afghanistan’s neighboring countries

Pakistan and Iran had wind power installed capacities of 106 MW and 91 MW in 2013, respectively (GWEC, 2014). Although there have been concerns about the consequences of high penetration of variable renewable energy sources, solar PV and wind power are achieving high levels of penetration in many countries. Denmark for example, met 30 % of its electricity demand with wind and 5.6 % with solar PV in 2012 (REN21, 2013).

Afghanistan is one of the least developed countries in the world located in south

Asia bordering Islamic Republic of Iran from west, Pakistan from south and east, China from northeast and Tajikistan, Uzbekistan, and Turkmenistan from North (Figure 1). It is located between latitude 29° 35’ and 38° 40’ degrees north and longitude 60° 31’ and 75°

00’ east. Its total land area is 652,864 km2 which is slightly smaller than the state of

Texas. 2012-2013 estimated population is 25,500,100 excluding temporary Afghan refugees living in Iran and Pakistan and only about 24 % of the population lives in urban areas (CSO, 2014). Afghanistan has an arid and semi-arid climate with rugged mountains and some plains in north and southwest (NOAA, 2008).

Afghan society has been very vulnerable and insecure over the past few decades.

It was ranked 175th on the United Nation’s Human Development Index and the lowest in

Asia in 2012 (UNDP, 2013). In 2008, 35 % of the population was unemployed and 36 % of them lived under poverty line (CIA, 2014). Literacy rate in 2010 was about 30 % and is expected to increase to 50 % by 2015 (ACCU, 2010). Gross national income (GNI) per

Afghan was $ 680 in 2012 (WB, n.d.). Foreign aid plays a major role in the economy.

Afghanistan received more than $ 56 Billion in development aid during 2002 and 2010 which translates to about 78 % of GDP since 2003 (Fichtner, 2013).

2

Figure 1. Map of Afghanistan 1 Gross electricity consumption in Afghanistan was 140 kWh per capita in 2011, one of the lowest rates in the world (Fichtner, 2013). Average household electricity consumption rate varies from 3000 kWh/year in Kabul Province to 178 kWh/year in

Ghor Province (Fichtner, 2013). On the other hand, gross electricity demand is growing at a staggering rate of 8.7 % per year as the country’s population, GDP and income levels are growing and political and economic stability is being achieved (Fichtner, 2013).

Annual gross demand for the whole country is expected to increase from 3,531 GWh

(2011) to 18,409 GWh (2032) and annual peak demand from 742 MW (2011) to 3,502

MW (2032) (Fichtner, 2013). This growth in demand means that Afghanistan will need

1 http://nzakariah18gi.blogspot.com/2012/08/background-to-breadwinner.html

3 about five times more electrical energy than was produced in 2011. Only 28 % of the population was connected to electricity grid and this number is projected to reach to about 83 % by 2032 (Fichtner, 2013).

Afghanistan so far does not have an interconnected centralized power system, however, interconnection of all grid segments is proposed by year 2032 (Fichtner, 2013).

In addition, there are many decentralized local grids and stand-alone systems such as solar PV and diesel generators providing electricity. Installed capacity (not operating capacity) of existing grid-connected electricity generation assets reaches about 1,338

MW including imports from Tajikistan, Uzbekistan, Iran and Turkmenistan. Imports account for about 63 % of total grid-connected capacity, while hydro power and thermal

(diesel-fired) power plants make up the rest each having about the same share. Currently about 134 MW of decentralized power generators are installed around the country mostly in rural areas, more than half of which is diesel generators. 13 MW of solar PV, 36.65

MW of micro-hydro power and about 200 KW of wind power make up the rest of Afghan decentralized generating capacity (MEW , 2013).

Figure 2 shows the share of each energy source in the Afghan power system in

2011. Total generation was recorded to be 3,088 GWh (AEIC, 2012). Imports made up about 73 % of total generation. The majority of imported power was thermal-based with some hydropower from Tajikistan. Domestic hydropower and diesel fired power plants each contributed 26 % and 1.3 % respectively. The share of diesel power plants could be higher due to missing and incomplete data.

4

Figure 2. Share of energy generation sources in 2011

A majority of existing and planned hydropower plants use day storage rather than run-of-the river generation (Fichtner, 2013). The main hydro plants with day storage are

Naghlu (100 MW) and Mahipar (66 MW). Examples of run-of-the river hydro plants are

Grishk (2.4 MW) and Puli-Khomri (8.2 MW). Thermal power plants are mainly reciprocating engines run by diesel fuel. The only two gas turbine generators are NW

Kabul 21.8 MW and NW Kabul 23 MW.

1.2. Literature Review

Stand-alone off-grid solar and wind energy technologies have thus far been seen as a favorable solution to electrify rural areas in Afghanistan by the government, donor agencies and communities themselves. The main reasons are lower operating costs, technical simplicity, and short installation time. Currently, there are no utility-scale solar

PV or wind power plants. The largest renewable energy system feeding a local grid is a 1

5

MW solar PV plant with battery storage in the central province of Bamyan. Here we review some of the main studies regarding the potential of large scale solar PV or wind power plants in Afghanistan.

NREL published a 1km resolution wind map at 50 m for Afghanistan in 2007 to quantify wind resource potential and identify possible locations for further on-site wind measurement campaigns (NREL, 2007). The dataset includes average monthly wind speeds together with statistical parameters such as wind power density, Weibull shape factor, windiest hour, diurnal strength, and the autocorrelation factor for each 1 km2 grid in Afghanistan. The above mentioned parameters are sufficient to calculate hourly wind speed for a site. However, wind direction, turbulence, and temperature are not included.

This dataset is not sufficient to model the energy production of a wind farm since wind speed varies over small distances and terrains. The accuracy of modeled data is in the range of ±20% due to lack of extensive field data as inputs to the model and

Afghanistan’s complex terrain. NREL’s estimates showed a total resource potential of about 158 GW.

Relative to solar energy, NREL also produced high-resolution satellite-derived global horizontal and direct normal irradiance data for Afghanistan (NREL, 2007). The datasets are in a gridded format and have a ground resolution of 0.1 degrees latitude and

0.1 degrees longitude (8.5 km x 10 km). Irradiance datasets include monthly averages over a three year period from 2002 to 2005.

In another study, Tetra Tech used multi-criteria geographic information system

(GIS) analysis to identify the best sites for future installation of wind measurement

6 instruments for wind farm development and their respective installed capacities and capacity factors (TETRA TECH, 2009). The study identified a total of 10 sites, located in

Herat (3), Balkh (5) and Kabul provinces (2). The study used NREL’s modeled wind resource maps, available GIS data, and measured wind speed data from Panjsher Valley to determine the technical and economic potential of these sites. Calculations are performed on an annual basis and do not study seasonal and daily characteristics of wind resource. In addition, the study does not cover demand characteristics of the provinces and whether the wind generation profile matches the demand profile. The total economic potential of these sites was estimated to be 1.045 GW of installed capacity.

NRECA International evaluated the response of transmission networks in Balkh and Herat provinces to carry hypothetical levels of wind penetration and determined the maximum installed capacity in Mega Watts (MW) of wind farms without disrupting the quality of the supplied power to the customers (NRECA , 2010). The reason for assuming hypothetical wind farm capacities was lack of wind resource characterization and daily profiles. The demand profile of Kabul was used to model daily demand profiles in Herat and Balkh provinces. Criteria for the determination of the maximum capacity of wind farms were an installed capacity of no more than 80 % of the minimum grid demand, voltage stability at specified levels during short-term wind farm output fluctuations, prevention of overloaded facilities and respecting limits of reactive power at all generators. Results showed that Balkh and Herat power systems could handle wind power plant installed capacities of up to 85 MW and 16.5 MW respectively if these systems were to be upgraded.

7

Foster assessed the feasibility of interconnecting a 10 MWe photovoltaic (PV) system in the Kandahar City utility as part of USAID funded Afghanistan Clean Energy

Program (ACEP) (ACEP, 2010). This study assumed a 1-axis tracking PV system without storage connected to the local grid. Monthly solar radiation and air temperature were used to calculate energy generation and the capacity factor for the system. A life

Cycle Cost (LCC) analysis were used to determine total installed cost, net present value and amortized cost of energy. The annual average estimated capacity factor was 23 %.

Foster also compared a proposed 1 MW grid-tied solar PV power system with a potential 1 MW micro-hydro power scheme to electrify communities in the province of

Bamyan (ACEP , 2011). The solar PV system was proposed by the New Zealand

Provincial Reconstruction Team (PRT). He concluded that it is a logical choice to install a 1 MW micro-hydro power plant rather than a 1 MW PV system due to lower capital cost, lower cost of energy per kWh, and production and demand profile matching for the particular communities in Bamyan. Annual average estimated capacity factor for solar

PV was estimated to be about 13 %.

Ershad estimated Afghan resource potential of solar PV to be about 30, 987 GW while its technical potential being about 29 GW (Ershad, 2014). The study used a multi- criteria GIS analysis of the solar resource together with topographic and geographic constraints. Future solar PV farms were assumed to be located within 40 km of five major demand centers (Kabul, Mazar-e-Sharif, Herat, Kandahar and Jalalabad) and within 2 km of an existing transmission line. Topographic constraints included lands with less than 3

% slope and land uses limited to only rangeland, bare soil and sand covered areas. No

8 constraint was applied for solar radiation. The study did not include economic or market potential analysis.

Finally, Fichtner concluded in Afghanistan Power Sector Master Plan (PSMP) that solar PV and wind power plants could not achieve high penetration levels in the existing and future power system and their role in the mix of grid-connected power generation is said to be minimal (Fichtner, 2013). Fichtner recommends developing distributed hybrid wind, solar and diesel power plants and off-grid solar home systems to meet the demand mainly in rural areas.

1.3. Research Objective

The purpose of this study is to analyze the potential of solar PV and wind power plants in Balk and Herat Provinces, two of the most promising provinces for future renewable power generation. Western Herat and Eastern Balkh are said to be two of the major wind resource areas in the country (NREL, 2007). In addition to wind, these provinces have excellent solar resource. Other factors that make these provinces very attractive to renewable energy deployment are their relatively high electricity demand, well developed power systems, heavy reliance on imported power and proximity to the western and northern borders.

The potential for solar PV and wind power plants is analyzed by quantifying resource quality, variability and cost of energy generation. Quality is quantified in section two by characterizing solar and wind energy resources and plant capacity factors and energy yields. Section three analyzes electricity demand and how it changes on a daily and seasonal basis.

9

Using data from sections two and three, variability (section four) is analyzed using three different methodologies. First, seasonal and diurnal profiles of power output from wind and solar PV power plants are correlated with temporal electricity demand profiles analyzed in section 4.1. Second, the impacts of power supply from solar PV and wind on various types of loads2 such as peak, cyclic and baseload are studied by creating residual load duration curves (RLDC) explained in section 4.2. Third, variability is measured in terms of curtailment and penetration. Curtailment is associated with times when the total capacity including that coming from renewable sources exceeds demand, and penetration refers to the fraction of total energy provided by renewable energy.

Section 4.3 describes our effort to estimate penetration and curtailment levels of three solar PV and wind power plant sizes (50 MW, 100 MW, and 150 MW) interconnected to grid segments with varying peak loads (200 MW, 300 MW, 400 MW and 500 MW).

Finally, cost of energy generated from solar PV and wind power plants is determined in section five and the thesis is concluded in section six.

2 Refer to (Ueckerdt, Brecha, & Luderer, 2014) in the References section for more detail on types of loads

10

2. RESOURCE QUALITY

2.1. Datasets

Historical measured wind and solar data is very limited in Afghanistan. In 2012, the ACEP funded by the USAID conducted a wind monitoring campaign to measure wind data for a period of 12 months for commercial applications at an appropriate height.

Six sites (Table 1) were selected using NREL’s modeled wind resource maps and other important criteria for wind farm siting such as proximity to transmission lines. Ten- minute interval wind speed and wind direction data at 30 m, 40m and 50 m elevations were collected for a period of one year in 2012 at each potential site. In addition, air temperature at 3 m and global horizontal solar radiation were also measured.

For the purpose of this study, the wind resource at Hotel Safid for Herat Province and Uljato for Balkh Province are characterized and later used for wind plant power output calculations. These two sites have the highest annual average wind speed among the monitored sites and are in the proximity of the sites selected for further measurement by a previous study (TETRA TECH, 2009).

11

Station Latitude Longitude Elevation Province Time Zone Location (°) (°) (m)

Hotel Safid Herat 34.4054 61.8226 958 + 4 30 Urdu Khan Herat 34.3126 62.2677 958 + 4 30 Uljato Balkh 36.7212 67.6222 385 + 4 30 Sari-Tangi Balkh 36.6278 67.6994 1012 + 4 30 Naghlu Kabul 34.6220 69.6959 1199 + 4 30 Jabal Seraj Parwan 35.1255 69.2348 1199 + 4 30

Table 1. ACEP wind monitoring sites in 2012

Collected data were validated by comparing them with allowable upper and lower limiting values called range tests (AWS Scientific, 1997). Table 2 shows ranges of two main parameters used in the test. No problems were observed with the data. Table 3 shows details of the dataset properties for Uljato and Hotel Safid.

Wind Speed Offset (0.35) < Avg < 25 m/s

Wind Direction 0 ° < Avg < 360 °

Table 2. Range test criteria

In addition, to verify the time step correctness of the datasets, measured temperatures were matched with the data from the airports. As expected, measured temperature peaked in the afternoon at both sites. Missing data for the period of

01/01/2012 to 01/16/2012 was filled with data from 01/01/2013-01/16/2013 and for the period of 02/01/2012 to 02/04/2012 with data from 01/30-01/31 and 02/05-02/06/2012 for the Hotel Safid site. Feb 29 is not included in the analysis. NRG tall towers were used

12 to measure the data. Table 4 shows the operating ranges and accuracy of measurement sensors.

Variable Uljato Hotel Safid Latitude N 36.7212 N 34.4054 Longitude E 67.62222 E 61.8226 Elevation 385 m 958 m Height of Measurement 50 m, 40m, 30m, 3m 50 m, 40m, 30m, 3m Start Date 1/1/2012 0:00 1/17/2012 0:00 End Date 1/1/2013 0:00 1/16/2013 0:00 Duration 12 months 12 months 2012 Data Recovered 99.73% (Including 2/29) 95.36 % (Including 2/29) Length of Time step 10 minutes 10 minutes Mean Temperature 18.7 °C 18.8 °C Mean Air Density 1.157 kg/m3 1.081 kg/m3

Table 3. Dataset properties for Uljato and Hotel Safid sites

Parameter Sensor Type Operating Range Accuracy NRG #40 Maximum ±0.14 m/s - Wind Speed @ 50 m Anemometer 1 m/s - 96 m/s ±0.45 m/s NRG #200P Wind 8° maximum Wind Direction @ 50 m Direction Vane 0° - 360° Temperature NRG#110 S -40 °C to 52.5 °C ±1.1 °C Global Horizontal Li-Cor #Li 200SA Irradiance Pyranometer 0 - 3000 W/m2 ±2 %

Table 4. Operating ranges and accuracy of measurement sensors used by ACEP

NREL’s Typical Meteorological Year (TMY) weather files for Mazar-e-Sharif

International Airport (OAMS) and Herat International Airport (OAHR) are used to calculate the energy production of solar PV power plants in Balkh and Herat. In addition, a separate spreadsheet is developed to calculate the output of a PV array in Balkh province using actual measured global horizontal irradiance (GHI) and ambient

13 temperature from the Uljato wind monitoring station for comparison reasons. Measured solar radiation data from Hotel Safid included errors due to the malfunctioning of installed Pyranometer and is not used in calculations.

2.2. Wind Resource Characteristics

2.2.1. Wind Speed

It is common to see different seasonal variations throughout the world and even a country. Data from 2012 shows that a spring (March) maximum and a fall minimum

(October) occur in Uljato. The windiest month in Hotel Safid occurs in August while

December experiences the lowest wind speed. Table 5 shows the annual average wind speed at 30 m, 40 m, and 50 m at both sites. Figures 3 and 4 show seasonal variation of wind speed together with standard deviation at 50 m in Uljato and Hotel Safid respectively. There is a bigger difference in monthly wind speeds between the three measurement heights in summer in Hotel Safid and winter in Uljato due to the higher wind speed standard deviations.

Uljato Hotel Safid Height (m) Wind speed (m/s) 30 5.75 8.60 40 6.05 8.84 50 6.34 9.11

Table 5. Uljato and Hotel Safid annual average measured wind speeds in 2012

14

Figure 3. Seasonal variation of wind speed and standard deviation in Uljato at 50 m height

Figure 4. Seasonal variation of wind speed and standard deviation in Hotel Safid at 50 m height

Sites with higher wind speed in the winter are ideal for Afghanistan since electricity demand is highest and hydropower generation is lowest. Uljato experiences windier winter season than Hotel Safid. Long-term wind speed data from Mazar-e-Sharif

Airport (OAMS) and Herat Airport (OAHR) show that summers are windier than

15 winters. Figures 5 and 6 show the difference between measured wind speeds in Uljato and Hotel Safid and data measured at their nearest airports. Annual average wind speed in Afghanistan is said to be 6.5 m/s with a maximum of 10.2 m/s and a minimum of 3.7 m/s (NREL, 2007).

Figure 5. Monthly wind speed in Uljato and OAMS

Figure 6. Monthly wind speed in Hotel Safid and OAHR

16

It is important to notice that the above graphs do not show a typical behavior of monthly variation at these sites since monthly average wind speeds change significantly from year to year. Figure 7 shows changes in 3 year average monthly variation of wind speed for Mazar-e-Sharif which is about 50 kilometers west of Uljato site. This location experiences a fairly constant speed with a minimum in fall.

Inter-annual variations in wind speed can have a huge impact in long-term energy production of wind turbines. Thus, it is important to understand and determine the long- term annual average wind speed in Uljato and Hotel Safid sites. It takes at least 5 years to come up with a reliable annual average wind speed. However, using one year data is enough to determine the long-term average wind speed within an accuracy of 10 % with a confidence level of 90 % (Manwell, McGowan, & Rogers, 2009).

Figure 7. Changes in monthly average wind speeds in OAMS for the period from 2011 to 2013

17

A typical diurnal variation in wind speed shows an increase during the day and a decrease during the night (Manwell, McGowan, & Rogers, 2009). However, wind speeds in Uljato and Hotel Safid do not seem to depend strongly on time of the day. As shown in

Figure 8, annual average variation of daily wind speed in Uljato is within 10 % of its peak while in Hotel Safid it is within 30 %. Uljato experiences multiple peaks while

Hotel Safid seem to peak at around 5 p.m.

Figure 8. Hotel Safid and Uljato annual average daily wind speed profile

Diurnal pattern strengths for these sites are calculated by fitting a cosine function to their annual average diurnal profiles. Equation 1 (Lambart, 2004) shows the fitted cosine function.

2휋 푣 = 푣̅ {1 + 훿 cos [( ) (𝑖 − 휑)]} For 𝑖 = 1, 2, … 24 … (1) 푖 24

Where 푣̅ = annual average wind speed (m/s)

훿 = Diurnal pattern strength (0-1) 휑 = Hour of peak wind

18

Using NREL’s HOMER software and measured time-series wind speed data at 50 meters, diurnal pattern strengths of 0.00711 and 0.0705 are obtained for Uljato and Hotel

Safid respectively. Typical measured values for US cities range from 0 to 0.4 (Lambart,

2004) and modeled value for all of Afghanistan is assumed to be 0.15 (NREL, 2007).

The lower the value the less wind speed depends on time of the day. According to NREL wind maps and airport data, 5 p.m. is the hour of peak wind in the most of the regions in the country while some western regions experience a peak at 3 a.m. Figures 9 and 10 show diurnal variation of wind speed in Uljato and Hotel Safid and at their nearest airports.

Figure 9. Daily profile of measured wind speed in Uljato and OAMS

19

Figure 10. Daily profile of measured wind speed in Hotel Safid and OAHR

With regard to seasons, typical diurnal variations are highest in summer and spring and lowest in winter (Manwell, McGowan, & Rogers, 2009). As shown in Figure

11 and 12, diurnal variation in wind speed is much higher during the months of April through October in Hotel Safid. Wind in Uljato doesn’t follow a specific seasonal pattern.

Figure 11. Seasonal variation of wind speed diurnal profile in Hotel Safid

20

Figure 12. Seasonal variation of wind speed diurnal profile in Uljato

To understand how wind speed in one hour depends on the wind speed in the previous hour, autocorrelation functions of the data are created. Figure 13 shows the autocorrelation function of the hourly wind speed data measured in Uljato and Hotel

Safid. The autocorrelation factor, a parameter showing whether the wind speed is random or dependent on the previous values, is calculated using Equation 2 (Lambert, 2004) as shown below

푁−1 ∑푖=1 (푣푖 − 푣̅) (푣푖+1 − 푣̅) 푟1 = 푁 2 … (2) ∑푖=1(푣푖 − 푣̅)

In the above equation 푟1 is the autocorrelation between two time series wind speed data separated by one hour, N is the number of time series data which is 8760, 푣̅ is annual average wind speed in m/s. Autocorrelation factors for Uljato and Hotel Safid are calculated to be 0.926 and 0.943 respectively. Typical values for sites with complex topography are 0.70 – 0.80 and 0.90-0.97 for sites with uniform topography (Lambert,

2004). Uljato and Hotel Safid both have uniform topography.

21

One could also notice the daily pattern of wind speeds at Hotel Safid and Uljato by looking at the autocorrelation function plot. Afternoons tend to be windier and wind speed is autocorrelated strongly at shorter lags but not significantly less correlated at longer ones. This characteristic of wind speed is very important in predicting the wind power plant output for grid management.

Diurnal variations could change from year to year. Although gross features of the diurnal cycle can be depicted by one year of data, other details such as the amplitude of oscillation of this cycle cannot be understood precisely (Manwell, McGowan, & Rogers,

2009).

Figure 13. Autocorrelation function of measured wind speed in Hotel Safid and Uljato

2.2.2. Turbulence Intensity (TI)

Important short-term wind speed variations are due to turbulence and gust.

Turbulence is the random changes in wind speed during an interval of no less than 1 second to ten minutes (Manwell, McGowan, & Rogers, 2009) . Turbulence and gusts

22 need to be quantified for a site for purposes such as turbine structural stability, control and operation issues. To quantify turbulence, the turbulence intensity (TI) of the site needs to be determined. Turbulence Intensity is the ratio of the standard deviation of wind speed to mean wind speed (Manwell, McGowan, & Rogers, 2009). Turbulence intensity affects a wind turbine’s mechanical and structural performance. Thus, wind turbines are designed to work best in a specific turbulence range. International Electrical

Commission (IEC) standard 61400-1 defines four turbulence categories as shown in

Table 6 based on mean turbulence intensities at wind speed of 15 m/s, and wind turbine manufacturers rate their devices accordingly.

Category Mean TI at 15 m/s

S >0.16

A 0.14-0.16

B 0.12-0.14

C 0-0.12

Table 6. IEC 61400-1 Turbulence categories 3rd edition

TI at 15 m/s of 7 % and 11 % are observed in Uljato and Hotel Safid respectively.

Thus, both sites fall under IEC category C. Highest recorded 10-min wind speeds are

31.1 m/s and 32.9 m/s in Uljato and Hotel Safid respectively. Both sites have different seasonal but similar daily profiles of turbulence intensity. Winds in summer months in

Uljato unlike Hotel Safid are more turbulent than winter months. On a daily basis, both sites experience turbulent winds during the day obviously due to the heating of the air

23 during the day. Figures 14 and 15 show seasonal and diurnal variations of turbulence intensity at both sites.

Figure 14. Seasonal variation of TI in Hotel Safid and Uljato

Figure 15. Daily variation of TI in Hotel Safid and Uljato

24

2.2.3. Wind Direction

Understanding prevailing wind directions and wind direction distribution at a site to be developed for wind power production is very critical since they affect energy production and wind turbine structural performance. Wind in Uljato blows from two main directions; east and west. About 26 % of the time it blows from east (90°) and 15 % of time from west (270 °) at both 40 m and 50 m heights. However, prevailing energy- producing direction varies with height. At 50 m height, the most frequent and the most energy-producing directions are the same (east). However, at 40 m, west is the most energy producing direction. Unlike Uljato, Hotel Safid experiences winds only from south-south west (202.5°-225°). The most frequent and the most energy producing directions are the same. Figures 16-17 and 18-19 show wind direction as % percent of total energy and % of total time in Uljato and Hotel Safid respectively.

Figure 16. Wind speed direction as percent of total energy in Uljato (left) Figure 17. Wind speed direction as percent of total time in Uljato (right)

25

Figure 18. Wind direction as percent of total energy in Hotel Safid (left) Figure 19. Wind direction as percent of time in Hotel Safi (right)

2.2.4. Wind Shear

Wind shear coefficient of a site is an important parameter because it impacts cyclic loading on turbine blades and energy calculations (Ray, Rogers, & McGowan,

2006). The actual vertical wind shear exponent is calculated using the power law

(Janardan & Nelson, 1994) as shown by Equation 3

푣2 log10 훼 = 푣1 … (3) 푧2 log10 푧1

Measured annual average velocities at 50 m, 40 m, and 30 m heights are used to determine the average wind shear coefficient representing these sites. Average wind shear coefficient is 0.2 for Uljato and 0.11 for Hotel Safid as shown in Table 7. These values are in the range of values American Wind Energy Association (AWEA) specified for sites with low bushes and few trees which represent actual site conditions at these

26 stations. Both sites experienced a higher wind shear coefficient during the night on a daily basis and during winter on a seasonal basis. Figures 20 and 21 show seasonal and diurnal variations of wind shear coefficient for both sites.

Hotel Uljato H (m) Safid V(m/s) 50 6.35 9.10 40 6.05 8.84 30 5.75 8.60 α50/40= 0.22 0.13 α40/30= 0.18 0.09 α50/30= 0.19 0.11

αave= 0.20 0.11

Table 7. Average wind shear coefficients in Uljato and Hotel Safid

Figure 20. Monthly variation of average wind shear coefficient

27

Figure 21. Daily variation of average wind shear coefficient

2.2.5. Wind Speed Distribution

To better understand the wind resource and estimate the energy production in

Hotel Safid and Uljato, wind speed distributions were created. Figure 22 and 23 show the actual and fitted Weibull frequency of occurrence of wind speed in Uljato and Hotel

Safid at 50 m.

Figure 22. Uljato wind speed frequency distribution at 50 m

28

Figure 23. Hotel Safid wind speed frequency distribution at 50 m

Windograph software is used to find the best fit to the histograms of wind speed distribution. Maximum likelihood algorithm as shown in Equations 4 and 5 (Seguro &

Lambert, 2000) was chosen in Windograph to calculate Weibull shape factor k and the scale factor c for both sites.

푛 푘 푛 ∑푖=1 푣푖 ln(푣푖) ∑푖=1 ln(푣푖) −1 푘 = ( 푛 푘 − ) … (4) ∑푖=1 푣푖 푛

푛 1 푐 = ( ∑ 푣𝑖푘) … (5) 푛 푖=1

Where vi is the wind speed and n is the number of nonzero wind speed data points. Table

8 shows the results of calculations for both sites. An average Weibull shape factor for

Afghanistan is modeled to be 1.7 and has a range of 1.1 to 2.54 (NREL, 2007).

29

Parameter Uljato Hotel Safid Weibull Shape Factor 1.51 1.76 Weibull Scale Factor 6.99 10.17 R-square 0.93625 0.83842

Table 8. Weibull parameters for measured wind speed at 50 m height

Surprisingly, the wind speed at these sites as seen in the above frequency graphs peak twice. This type of wind regime is not very typical but seen in places where summer and winter winds are very different (Burton, Sharp , Jenkins, & Bossanyi, 2001). This type of wind regime is better modeled by a bi-Weibull frequency distribution function with different scale and shape factors in two seasons. The shape factors k1 and k2 together with scale factors c1 and c2 and coefficients f1 and f2 are determined using

Equation 6 as shown below

(푘 ) (푘 ) 푣푖 1 푣푖 2 (푘1−1) −(( ) ) (푘2−1) −(( ) ) 푣푖 푘1 푐1 푣푖 푘2 푐2 퐹 = 푓1 ∗ ( ) ∗ ( ) ∗ 푒 +푓2 ∗ ( ) ∗ ( ) ∗ 푒 … (6) 푐1 푐1 푐2 푐2

The curve-fitting tool in Matlab was used to determine the parameters mentioned above to find the best fit for the wind speed distribution. Table 9 shows the calculated parameters with the R-squared of the regression. Figure 24 shows the measured distribution together with the corresponding bi-Weibull distribution.

30

Parameters Hotel Safid Uljato f1 0.39 0.74 f2 0.61 0.25 k1 2.18 2.39 c1 5.46 9.35 k2 3.65 2.41 c2 14.04 3.08 R-square 0.92 0.97 Improvement 9 % 3 %

Table 9. Bi-Weibull parameters for wind speeds at 50 m in Uljato and Hotel Safid

Figure 24. Actual and bi-Weibull wind speed distribution frequency

2.2.6. Wind Power Density

Using measured air temperatures, and calculated average wind shear coefficients, the wind power density at 30m, 40m, 50 m and 70 m turbine hub heights is calculated and presented in Table 10. According to NREL’s classification of wind power in

Afghanistan, the wind power density at 70 m in Uljato is considered “Good” and falls

31 under class 4 wind resource. Wind power densities at all heights in Hotel Safid are considered “Excellent” and fall under class 6 and 7. Wind class 4 and above are deemed appropriate for wind farm development (NREL, 2007).

Uljato Hotel Safid Height (m) WPD (W/m2) 30 273 682 40 317 741 50 357 798 70 426 879

Table 10. Hotel Safid and Uljato wind power densities at 30 m, 40 m and 50 m height

2.3. Wind Plant Output

Hourly energy yields (MWh/MW) of potential wind power plants are calculated by converting time-series wind speed data to power output of wind turbines. Only one wind speed measurement is used to characterize the plant output at Uljato and Hotel Safid sites. This is an appropriate assumption since areas surrounding these measurement towers are pretty flat with similar topography for multiple turbines to be installed and the size of these plants if ever built would not be very big.

Power output of the potential wind farms is modeled using 10 minute-interval wind speed, wind direction and air temperatures, together with the composite power curve of a 630 KW wind turbine with a rotor diameter of 48 m as shown in Figure 25.

The composite power curve is created using power curves from commercial pitch- regulated wind turbines in the range of 275 kW – 1 MW which are deemed feasible for installation in Afghanistan (TETRA TECH, 2009). Each turbine’s power output was

32 averaged over a 0.5 m/s wind speed bin and the averaging process was performed over a range of air densities. The turbine has cut-in, rated and cut-out wind speeds of 3 m/s, 15 m/s, and 25 m/s, respectively.

Figure 25. 630 KW wind turbine composite power curve

The process of converting time-series wind speed into power output starts with calculating the wind speed at the desired turbine hub height. Output at 70 m hub height is calculated for Uljato and Hotel Safid wind farms to increase energy capture. Wind speed at 50 m height is converted to wind speeds at 70 m using power law (Equation 3) with wind shear coefficients of 0.2 and 0.1 for Uljato and Hotel Safid respectively.

ℎ2 훼 푣 = 푣 ∗ [ ] 2 1 ℎ1

Where v1 and v2 are wind speeds at 50 m and 70 m, h1 =50 m and h2 =70 m and α is the wind shear coefficient.

Next, wind speed is adjusted to account for the difference in local and standard air densities. This adjustment is necessary when using turbine power curves developed for

33 air density at seal level of ρo = 1.22g kg/m3. Equation 7 (AWS Scientific, 1997) is used to calculate 10-min average air density for the wind farm sites from measured air temperature assuming a standard sea level air pressure.

푧 353.05 ( )∗−0.034 3 휌10 푚푖푛 = ( ) 푒푥푝 푇 … (7) (Kg/m ) 푇10 푚푖푛

Where: z is the site elevation above sea level (m)

T is 10-min average air temperature (K)

Equation 8 (Wan, Ela, & Orwig, 2010) is used to adjust the wind speeds at 70 m hub height for air density

1 휌10 푚푖푛 3 푣 = 푣10 푚푖푛 ∗ ( ) … (8) 휌0

Where:

휌10 푚푖푛 is the 10 min-average air density calculated from measured air temperature

푣10 푚푖푛 is the 10 min-average measured wind speed in m/s

3 휌0 is the sea level air density equal to 1.225 kg/m

Annual average wind speed at 70 m is reduced about 1.8 % in Uljato and 4.5 % in

Hotel Safid when adjusted for local air density. Higher loss in Hotel Safid is due to its higher elevation.

34

Next, the wind speed is adjusted for wake losses as a function of wind direction relative to the prevailing wind direction. This method is developed by AWS Truewind based on actual data from wind farms installed in the United States. Although, wake losses very much depend on wind farm layout, local topography and wind regimes, these estimates will be appropriate for Uljato and Hotel Safid farms. Equation 9 (AWS

Truewind, 2009) is used to calculate wake loss for each 10 min-interval wind speed.

2 푤10 푚푖푛 = 푤푚푖푛 + (푤푚푎푥 − 푤푚푖푛) ∗ sin(휃10 푚푖푛 − 휃푚푎푥) … (9)

Where:

푤푚푖푛 is the minimum assumed wake loss when the wind direction is aligned or opposite to the prevailing wind direction. It is assumed to be 4 %.

푤푚푎푥 is the maximum assumed wake loss when the wind direction is perpendicular to the prevailing wind direction. It is assumed to be 9 %.

휃10 푚푖푛 is the 10 min average measured wind direction

휃푚푎푥 is the prevailing wind direction from the north direction. It is 90 degrees in Uljato and 225 degrees in Hotel Safid.

Loss factors estimated using the above equation also accounts for such losses as blade soiling and icing. In addition, the above equation is developed to estimate wake losses in sites where there is only one prevailing wind direction or the most frequent wind direction is the same as the most energy-producing one. In Uljato the most frequent wind direction (east) is opposite to the most energy producing direction (west). In Hotel Safid, both directions are the same (southwest). Average wake loss for both sites is about 5 %.

35

To convert the air density and wake adjusted time-series wind speeds into 10- mintue average power yield in MWh/MW of wind farm installed capacity, the composite power curve of the 630 KW wind turbine is converted to a continuous function of wind speed. A 3rd degree polynomial curve was fitted to the binned data for the region from the cut-in wind speed of 3 m/s to the rated wind speed of 15 m/s (Figure 26). The fitted equation gives power values more than 630 KW in the range of 13 m/s to 14.5 m/s. To avoid this, the rated wind speed is moved to 13 m/s instead of 15 m/s. The following piece-wise equation (Equation 10) is obtained with a R2 of 99 %.

0 푣 < 3 푚/푠 푚 −1.2465 ∗ (푣)^3 + 31.956 ∗ (푣)^2 − 178.46 ∗ (푣) + 297.86 3 ≤ 푣 < 13 푚/푠 푃(푣) = 푠 … (10) 630 13 푚/푠 ≥ 푣 ≥ 25 푚/푠

{ 0 푣 > 25 푚/푠

Figure 26. Actual and fitted power curve of the 630KW wind turbine

Losses associated with increase in turbulence intensity from the reference turbulence intensity of the turbine power curve is not quantified since the information on

36 range of turbulence intensities of the power curve is not available and both sites have relatively low annual average turbulence intensities. Other losses such as electrical losses and wind turbine and grid unavailability losses are applied to the final power output.

Table 11 summarizes calculated and assumed losses for Uljato and Hotel Safid wind farms.

Loss Category Uljato Hotel Safid

Wake effects 5.0% 4.4%

Density adjustment 2.0% 4.2%

Electrical3 3.0% 3.0%

Availability4 13.0% 13.0%

Total 21.43% 22.7%

Table 11. Calculated and assumed wind farm losses for Hotel Safid and Uljato

Final results include hourly average power output and annual average net AC wind power and net capacity factors for Hotel Safid and Uljato wind farms. Table 12 shows net annual energy production (AEP) and net capacity factor for both sites. Table

12 also gives typical annual capacity factors of wind power plants around the world (IEA,

2013). Figure 27 and 28 show seasonal variation of net capacity factors and annual average daily net energy yield of both wind farms.

3 “Development of Eastern Regional Wind Resource and Wind Plant Output Datasets” by AWS Truewind 2010 4 “Development of Wind Energy Meteorology and Engineering for Siting and Design of Wind Energy Projects in Afghanistan” by TETRA TECH 2009

37

Site AEP (MWh) CF (%)

Hotel Safid 3,709 42.4

Uljato 2,418 27.6

World Typical 1,752-3,066 20-35

Table 12. Net annual energy production and capacity factors for Hotel Safid and Uljato wind power plants and typical capacity factor around the world

Figure 27. Monthly net capacity factors of Hotel Safid and Uljato win power plants

Figure 28. Annual average daily net energy yield of Hotel Safid and Uljato wind power plants

38

2.4. Solar Resource Characteristics

Afghanistan receives on average about 5.3 kWh of solar radiation per square meter of horizontal surface on a clear day with a standard deviation of 0.42 kWh. This corresponds to an annual average Global Horizontal Irradiance (GHI) of 1,935 kWh/m2.

National average seasonal maximum and minimum GHI are 7.84 kWh/m2/day and 2.38 kWh/m2/day. Figure 29 shows seasonal variation of GHI for Afghanistan (NREL, 2007).

Annual GHI for Herat and Balkh provinces are 1726 kWh/m2 and 1967 kWh/m2 respectively. Figure 30 shows resource map of GHI for Afghanistan.

Figure 29. Monthly global horizontal irradiance (GHI) in Afghanistan

Tables 13 and 14 show daily and annual average GHI data for OAMS and OAHR.

In addition, for comparison reasons, measured GHI from the Uljato site, which is about

40 km east of OAMS, is also given. Modeled data agree well with measured data.

39

Figure 30. Map of annual global horizontal irradiance (GHI) for Afghanistan

SOLAR RADIATION SUNY Model NASA SSE Model NREL CSR Model NREL TMY Data Uljato Measured Data Daily Average (kWh/m2/day) 4.69 4.83 4.86 4.82 4.80

Annual Average (kWh/m2/year) 1,713 1,763 1,773 1,759 1,751

Table 13. OAMS annual and daily average GHI

SUNY Model NASA SSE Model NREL CSR Model HERAT TMY SOLAR RADIATION (kWh/m2/day) (kWh/m2/day) (kWh/m2/day) (kWh/m2/day) Daily Average (kWh/m2/day) 5.378 5.041 5.520 5.383 Annual Average (kWh/m2/year) 1,963 1,840 2,015 1,965

Table 14. OAHR annual and daily average GHI

Figures 31 and 32 show seasonal variation of GHI for OAMS and OAHR by available datasets. Figure 33 depicts typical daily solar radiation profile on a horizontal

40 surface in Herat (OAHR) and Mazar-e-Sharif (OAMS) International Airports. 740 W/m2 and 664 W/m2 are annual average peaks in OAHR and OAMS at noon.

Figure 31. Annual average monthly GHI at OAMS by various models

Figure 32. Annual average monthly GHI at OAHR by various models

41

Figure 33. Annual average daily global horizontal irradiance at OAMS and OAHR

2.5. Solar PV Plant Output

NREL’s System Advisor Model (SAM 9.20.2013) software is used to calculate

PV array outputs. Hourly gross DC (Pgdc) output (accounting for effects of cell temperature) of a fixed 1000 KWdc PV array made up of 250 W Sharp NU-Q250W4 panels with a 15.3 % efficiency facing south tilted at an angle equal to the latitude of

OAHR (34.217 ° N) and OAMS (36.7 ° N) is calculated using TMY datasets from

OAMS and OAHR.

Pgdc in KWdc is then converted to net AC (Pacn) output after applying an overall derate factor of 77 % which is NREL’s old (before 2014) default value for grid connected

PV systems with crystalline modules and is a good approximate for conditions in

Afghanistan. Loss from shading is not included. Derate factor account for PV module nameplate derate, mismatch of PV modules in the array, DC and AC wiring losses, diodes and connections, soiling, availability and interconnection losses such as inverter and transformer. For a detailed proportion of losses from each category, refer to PVWatts

42 website5. More details on how SAM calculates incident radiation on the plane of array from horizontal radiation and cell temperature to account for temperature coefficient of power are given in the Help section of the software.

Net AC output of a PV array using measured hourly GHI and ambient temperature in 2012 from Uljato is calculated for a similar fixed 1000 KW dc PV system installed at latitude tilt (36.7° N) facing south using procedures adopted by NREL’s

PVWatts hourly simulation model. Hourly incident radiation on the tilted surface from measured global radiation is calculated using Homer’s time-series import option. Homer uses algorithms developed by Erbs, Klein and Duffie (1982) to estimate the horizontal diffuse and beam portions of the GHI to model plane of the array radiation.

Next, the cell temperature is calculated using Equation 11 (Messenger & Ventre,

2003) to account for DC power loss or gain due to an increase or decrease in temperature of PV cells from the reference cell temperature at Standard Test Conditions (STC) of 25

°C. The temperature coefficient of power is taken -0.5 % /°C.

푇푁푂퐶푇 – 20 푇 = 푇 + ( ) ∗ 퐺푡 … (11) 푐푒푙푙 푎 0.8

Where

Gt = the incident solar radiation on the tilted surface of the array (W/m2)

Ta = the ambient temperature in °C

TNOCT = the nominal operating cell temperature of the PV module and is 49 °C for most

PV modules

5 http://pvwatts.nrel.gov/pvwatts.php

43

Cell temperatures are then used to calculate gross DC output (Pgdc) using the

Equations 12 and 13 as shown below

퐺 푊 푃 = 푡 ∗ 푃 ∗ (1 + Ƴ (푇 – 푇 )) 푓표푟 퐺푡 > 125 … (12) 푔푑푐 1000 푑푐0 푐푒푙푙 푟푒푓 푚2

0.008 푊 푃 = ∗ 퐺 2 ∗ 푃 ∗ (1 + Ƴ (푇 – 푇 )) 푓표푟 퐺푡 ≤ 125 … (13) 푔푑푐 1000 푡 푑푐0 푐푒푙푙 푟푒푓 푚2

Where

Pdco = the installed DC capacity (KW)

Ƴ = the temperature coefficient for power taken -0.5 %/°C

Tcell = the cell temperature in °C

Tref =standard test condition (STC) reference temperature of 25 °C

To calculate net AC output, an overall derate factor of 77 % is applied to gross

DC output (Pgdc). Further procedures include decomposing the overall derate factor to pre-inverter and inverter losses to account for inverter part-load efficiencies. These steps are explained in Help section of PVWatts Calculator6. Table 15 shows annual net energy yields and net capacity factors for solar PV power plants. Each system requires about

16,000 m2 of land. Table 15 also gives typical annual capacity factors of solar PV power plants around the world (IEA, 2013).

6 http://pvwatts.nrel.gov/pvwatts.php

44

Site AEP (MWh/MW) CF (%) OAHR 1,482 17 OAMS 1,339 15.3 Uljato 1,260 14.4 World Typical 876-1,752 10-20

Table 15. Net annual energy production and capacity factors for Hotel Safid and Uljato solar PV power plants and typical capacity factor around the world

The PV array in Herat produces about 10 % higher energy than the one in Balkh due to higher solar resource. The difference in energy production of Balkh PV arrays could be due to modeling approaches like the variable inverter efficiency. Figures 34 and

35 show monthly net capacity factors and annual average diurnal AC output of the PV arrays respectively. Maximum energy yield at noon in Herat and Balkh are determined to be 0.57 MWh/MW and 0.51 MWh/MW of installed capacity respectively.

Figure 34. Monthly net capacity factor for solar PV power plants in Balkh and Herat

45

Figure 35. Annual average diurnal AC power output of solar PV power plants in Balkh and Herat

46

3. ELECTRICITY DEMAND

The only available record of time-dependent electricity consumption in

Afghanistan is the measured gross power production of domestic plants and import lines from 2006 through 2011 (AEIC, 2012). Balkh demand was served by imported power from Uzbekistan and Herat demand was served by power imported from Turkmenistan and Iran. The power network in Balkh province is part of North East Power System

(NEPS) that also serves many other provinces including Kabul in an islanded fashion.

Kabul demand for example is served both by imports from Uzbekistan and domestic hydro and thermal power plants. Herat on the other hand has its own power system.

For the purpose of this study, it is assumed that the proposed solar and wind farms in Balkh and Herat Provinces will be connected to grids with similar temporal load characteristics as the one in Kabul. However, there exist some obvious consumptions differences between Kabul, Balkh and Herat, such as higher load factors in Herat and

Balkh provinces due to greater industrial activity and higher and longer summer demand due to their higher annual average temperatures and thus increased use of air conditioning. In spite of this, Kabul temporal demand could be a reasonable proxy for demand in other provinces for two main reasons. First, it takes into account generation from domestic hydro and thermal sources. Second, all of these provinces would be

47 interconnected in the future and the single demand profile would be created dominated by

Kabul consumption – as Kabul is by far the most populous region in the country.

An Average Demand Profile (ADP) is created using Kabul demand data from

2010 and 2011 to be used in the analyses in section 4. Hourly demand data from each year are normalized to the peak value and averaged over both years for all 8,760 hours in a year. Thus, for a given peak demand, the ADP is scaled to obtain the corresponding hourly demand data. The minimum load of ADP is 19% of its peak demand. A gross load factor (since demand data used is in the form of gross power plant production) of 55% is achieved when using the ADP as a demand profile. To better understand ADP, temporal characteristics and other parameters of Kabul demand during 2010 and 2011 is explained in sections 3.1 and 3.2.

3.1. Seasonal Profile

Table 16 shows annual average daily energy, maximum and minimum power, and load factor for Kabul. The reason for a reduction in load factor in 2011 is an increase in the capacity of imported power in winter of 2011 which is also the period of highest demand. It is important to mention that these load factors are calculated from load data that do not necessarily meet 100 % of actual peak demands. Served demand is reported to be around 90 % of actual peak demand (Fichtner, 2013).

48

Parameter 2011 2010 Average (MWh/d) 4,249 3,345 Average (MW) 177 139 Peak (MW) 335 248 Minimum (MW) 20 6 Load Factor 52.90% 56.20%

Table 16. Kabul Province demand statistics and load factor in 2010 and 2011

Figure 36 shows a load histogram for both years. 50-60 % of the time the load was between 25 to 30 % of the peak for both years. Loads less than 10 % of the annual peak happen only for 1 hour in 2010 and 2 hours in 2011.

Figure 36. Load histogram for 2010 and 2011

Average 2010 gross electricity demand in winter (October-March) in Kabul was

35 % higher than summer (April-September) demand. This seasonal variation in demand is validated with 2011 demand and a 5 % increase in noticed. One of the main reasons for higher winter electric consumption could be electric heating. Summer demand peaks in

49

August and is about 63 % of overall peak. Summer peaks in Herat and Balkh are about 82

% and 87 % of annual peak respectively. Figure 37 shows annual average monthly demand for 2010 and 2011 in Kabul. The month with the highest annual average demand was December both in 2010 and 2011. Lowest electricity consumption is noticed during the months of May and September in Kabul with May being the overall minimum.

Figure 37. Kabul province average monthly demand in 2010 and 2011

3.2. Diurnal Profile

The load profile in Kabul Province is bimodal. This characteristics is observed thoughout the year. On average the hours with the highest electricity demand in 2011 were 8 a.m. and 8 p.m.. The hour with the lowest demand was 2 a.m. Figure 38 shows

2010 and 2011 normalized daily load profiles for Kabul. Demand dependancy on day of the week is assumed to be minimal since electricity is consumed mosty for ligting, cooking and heating as seen from the time of daily peak. Minimum average daily demand in 2011 was about 53 % of daily peak. This figure was about 47 % in 2010. Minimum

50 daily load is higher in Herat and Balkh provinces since these provinces have higher load factors.

Figure 38. Kabul 2010 - 2011 annual average daily demand profile

51

4. RESOURCE VARIABILITY

The capacity of wind and solar power plants depends highly on the correlation of availability of wind and solar with power demand, flexibility of other units in the system, available storage and interconnection and the share of variable renewables. The goal of this section is to quantify the variability of wind and solar resources in Balkh and Herat provinces by 1) comparing their daily and seasonal profiles with ADP 2) analyzing their impacts on load duration curves and 3) determining their penetration and curtailment levels under various peak demands using ADP.

4.1. Seasonal and Daily Correlation

As noted in the previous section, demand for electricity is higher in winter

(October through March) than summer (April through September). This period of high demand does not coincide with the strongest period of the wind resource in Herat or with peak solar radiation anywhere in the country. However, the wind resource in Uljato for the months of October through March is higher than April through September which closely matches the demand. In addition to wind and solar resources, the Afghan hydro resource also peaks in the spring and early summer months. On average, existing and future hydropower plants produce 100 of their rated power capacity in the months of May and June with the lowest production coming in January (Fichtner, 2013).

52

Figures 39 and 40 show seasonal variations of potential wind and solar PV power plants together with electricity demand and hydro power plants. Seasonal production profile (% of nameplate capacity) of Salma (Herat) and Mahipar (Kabul) hydropower plants are compared with solar and wind power profiles in Balkh and Herat respectively.

Of the three power plant types, solar PV experiences the least seasonal variation, while hydro experiences the most. This is true for Herat as well. Hotel Safid wind power complements Salma hydropower supply pretty well since hydro power peaks during

March through June, while wind power peaks during June through August.

On a daily basis, electricity demand is highest in the evenings with another lower peak in the mornings. Hotel Safid and Uljato wind farms would produce more power in the mornings than evenings during the months of November through March. During summer, a period of high wind power occurs during mid-day through the evening. This difference in winter and summer winds causes the wind resource to not depend on time of day during a year. Power output of solar PV plants plays little role in meeting peak daily demand since the peak demand occurs during the evenings. Figure 41 and 42 depicts daily profiles of demand versus power production of solar PV and wind plants.

53

Figure 39. Seasonal variation of intermittent renewables and electricity demand in Balkh

Figure 40. Seasonal variation of intermittent renewables and electricity demand in Herat

54

Figure 41. Annual average daily profile of demand versus wind and solar in Herat

Figure 42. Annual average daily profile of demand versus wind and solar in Balkh

4.2. Residual Load Duration Curves

The type and geographic location of intermittent power sources could have different impacts on the remaining load requirements. To understand the effects of Balkh and Herat solar PV and wind power on ADP, the method of Residual Load Duration

Curves (RLDC) (Ueckerdt, Brecha, & Luderer, 2014) is used. This method simply subtracts hourly power produced from solar PV or wind power plants or a combination of

55 them from the required demand for that hour to get the residual load. The resulting residual load is then sorted from highest to lowest to obtain the residual load duration curve. One can easily understand the temporal impact of renewables on the demand profile by comparing load duration curves (LDC) with RLDC. Although Herat wind and solar PV power plants achieve higher annual average capacity factors or number of full load hours than the ones in Balkh, their contribution in meeting various types of load such as peak, cyclic (intermediate) or baseload differ. In both locations, solar power plants reduce cyclic loads but not afternoon minimum loads or morning and evening peaks loads. This is true for Herat solar plants as well.

Uljato wind reduces all three types of loads. It has the same effect as solar PV for cyclic loads. The greatest contribution of Uljato wind is reducing operational hours of near peak load. Herat wind, on the other hand, has a significant impact on reducing cyclic and base loads since it is strongest in the summer and in the afternoons and evenings.

Figures 43 and 44 show residual load duration curves of ADP for 300 MW peak after introducing 50 MW solar PV power and 50 MW wind power.

If synergistic effects of solar PV and wind are considered, 50 MW solar and 50

MW wind in Herat would have the same effect as 50 MW wind and 50 MW solar in

Balkh in reducing near peak loads. However, Herat combined solar and wind would contribute more in reducing the baseload. Figure 45 shows residual load duration curves of ADP for 300 MW peak demand in Herat and Balkh when both sources are connected to the grid.

56

Figure 43. Residual load duration curves (RLDC) of Herat grid after introducing solar PV and wind power independently

Figure 44. Residual load duration curve (RLDC) of Balkh grid after introducing solar PV and wind power independently

57

Figure 45. Residual load duration curves (RLDC) of Balkh and Herat grids after introducing solar PV and wind power together

4.3. Penetration and Curtailment Levels

Wind or solar PV penetration is the ratio of total annual utilized power from these power plants divided by total annual demand of the grid. Curtailment on the other hand is the ratio of total annual wind or solar power that is not used, divided by the total annual wind or solar PV power generated. Hourly utilized power is calculated by subtracting solar and wind power from replaceable demand (actual demand minus minimum demand or base load7). Hourly curtailment is calculated using the same procedure as utilized power whenever production from solar or wind is higher than replaceable demand. In the meantime, for a given demand, penetration increases as the installed capacity increases.

In this section, penetration and curtailment curves of three sizes of wind and solar PV power plants (50 MW, 100 MW, and 150 MW) are created for four ADP peak demand

7 It is assumed that renewables can only replace up to 81 % of the load at any time. It is technically and economically infeasible to turn on and off some power plants to accommodate more renewables.

58

(200 MW, 300 MW, 400 MW and 500MW). Different grid demands are considered due to the growing demand of grid segments in Herat and Balkh. Grid demand values are roughly based on estimated future values for Herat and Balkh power systems (Fichtner,

2013). Results are presented in Figures 46 through 53.

For example, one could look at Figures 46 and 47 and notice that Balkh wind power plant achieves penetration levels of 12 % to 30 % for the range of installed capacities at 200 MW peak demand and decreases almost linearly to 5 % to 15 % at 500

MW peak. The highest curtailment loss of about 23 % occurs when installing 150 MW wind power plants when the peak demand is 200 MW. For the rest of the combinations, curtailment losses are about 10 % or less.

Wind power plants would achieve higher penetration levels than solar PV in

Balkh and Herat Provinces for a given installed capacity and grid demand. Both solar PV and wind power plants in Herat achieve higher penetration levels than the ones in Balkh.

Although wind power plants could achieve higher penetrations levels than solar PV, they have higher curtailment losses. This fact is due to mismatch of wind resource with demand especially during the night times. Wind power curtailment losses are higher in

Hotel Safid than Uljato.

It is unlikely to see a cumulative installed capacity of solar PV and wind turbines in the near future that would surpass limits of significant curtailment losses. However, it is wise to take advantage of the modular nature of solar PV and wind turbines. Installed capacity could be increased in parallel with increase in demand or export opportunities.

In addition, excess power from solar PV and wind power plants could be used to

59 complement hydropower plants especially during summer months by pumping back water to the reservoirs during the night when the load is minimal.

Figure 46. Balkh wind power penetration

Figure 47. Balkh wind power curtailment losses

60

Figure 48. Balkh solar PV power penetration

Figure 49. Balkh solar PV power curtailment

61

Figure 50. Herat wind power penetration

Figure 51. Herat wind power curtailment losses

62

Figure 52. Herat solar PV power penetration

Figure 53. Herat solar PV power curtailment losses

63

5. COST OF ENERGY GENERATION

Zero fuel cost is one of the main advantages of power generation from renewable sources such as solar and wind. However, higher installation capital costs are one of the barriers in deploying these power plants. The good news is that the cost of wind turbines and PV modules are going down as these systems are installed more and more. The goal of this section is to estimate power generation cost in USD/kWh and AFG per kWh assuming an exchange rate of 48 AFS/USD from solar PV and wind power plants in

Balkh and Herat provinces. To make calculations simple, the cost of energy generation for a 1 MW power plant is calculated and it is assumed that generation over the life of the projects remains the same as the first year generation. In addition, no replacement cost and inflation is considered. Table 17 shows economic parameters 8 used in this section.

Net capacity factors and net annual energy yields for the power plants in Balkh and Herat are summarized in Table 18.

8 Parameters are taken from Afghanistan’s Power Sector Master Plan prepared by Fichtner, 2013.

64

Wind Installed Cost $/KW 2300 Wind Operating Cost $/KW/Year 81 Solar PV Installed Cost $/KW 3500 Solar PV Operating Cost $/KW/Year 63 Useful Life Years 20 Discount Rate 12%

Table 17. Economic parameters used in calculation of electricity generation cost of solar PV and wind power plants in Balkh and Herat provinces

AEP Power Plant Capacity Factor (MWh/MW) Herat Solar PV 16.9% 1480 Herat Wind 42.3% 3705 Balkh Solar PV 14.0% 1226 Balkh Wind 27.6% 2418

Table 18. Net annual energy production and capacity factors of solar PV and wind power plants in Balkh and Herat

Energy generation cost, the levelised cost of electricity (LCOE), is estimated using Equation 14 as shown below,

퐶푇표푡푎푙 퐿퐶푂퐸 = … (14) 퐸푇표푡푎푙 Where

퐶푇표푡푎푙 = total power plant cost (excluding interconnection or transmission costs) plus the cost of operating the power plants over the project lifetime

퐸푇표푡푎푙 = total generated energy over the project lifetime

Equation 15 is used to calculate total power plant cost (퐶푇표푡푎푙 ).

65

퐶푇표푡푎푙 = 퐶푀푊 + 푃푉퐹 ∗ 퐶푂&푀 … (15) Where

퐶푀푊 = Power plant installed cost

퐶푂&푀 = Power plant operating cost

푃푉퐹 = Present value factor

Annual operating cost is discounted to find the present value. Equation 16 as shown below is used to estimate the present value factor (PVF) of annual operating costs

1 − (1 + 훼)−푛 푃푉퐹 = … (16) 훼 Where

훼 = Discount rate (12 %)

푛 = project lifetime (20 years)

Total generated energy 퐸푇표푡푎푙 is calculated simply by multiplying Net Energy

Yield given in Table 18 by the project lifetime of 20 years. Table 19 shows the cost of energy for each power plant in Balkh and Herat Provinces. The cheapest energy is generated from the Herat wind power plant while the most expensive is from the Balkh

Solar PV plant.

66

LCOE COE Power Plant (USD/kWh) (AFG/kWh) Herat Solar PV 0.134 6.433 Herat Wind 0.039 1.877 Balkh Solar PV 0.157 7.554 Balkh Wind 0.060 2.880

Table 19. Electricity generation cost of solar PV and wind power plants in Balkh and Herat

Optimistic estimates show that generation cost of electricity in Afghanistan would reach 0.068 USD/kWh by 2032 (Fichtner, 2013). Calculations are based on power generation from coal, natural gas, diesel and imports. Thus, wind power plants have a greater potential to reach grid parity than solar PV. However, with the help of appropriate policies and declining system costs, both sources could become economically viable.

67

6. CONCLUSION

Afghanistan unfortunately has not yet taken advantage of its vast resources of renewable energies such as solar radiation and wind on a scale large enough to meet a significant fraction of its electricity demand. Two of the most promising provinces for renewable energy deployment are Balkh and Herat provinces. Annual average global horizontal solar radiation in Herat and Balkh are 1,726 kWh/m2 and 1,967 kWh/m2 respectively. Fixed axis solar PV power plants tilted at angles equal to the latitudes achieve 14 % and 17 % annual net capacity factors including modest losses in Balkh and

Herat respectively. Typical global capacity factors of solar PV power plants are in the range of 10 % to 20%.

Average wind speeds at 50 m and power densities at 70 m in Uljato and Hotel

Safid reach 6.34 m/s and 9.11 m/s and 426 W/m2 and 879 W/m2 respectively. Sites with power densities greater than 400 W/m2 are considered commercially viable for wind farm development. Wind power plants in Uljato and Hotel Safid yield annual net capacity factors of 27.6 % and 42.3 % respectively. Typical onshore wind power plants around the world achieve capacity factors in the range of 20 % to 35 % with exceptional plants achieving 45 %.

68

Electricity demand in Afghanistan is highest in the winter while power supplied from wind power plants in most of locations and solar PV power plants everywhere would be highest in the summer. Uljato is one of the locations that experiences stronger wind speeds in winter months (November – March). Hotel Safid on the other hand experiences higher winds in summer. Wind speed diurnal profile varies from month to month and it turns out that on average wind speed in both locations do not quite depend on time of day. Solar PV and wind power plants in Balkh and Herat have the potential to reduce the load on all types of domestic generation schemes and imports. For example, solar PV can reduce the number of operating hours of power plants responsible for providing cyclic power especially to meet mid-day demands. Wind power plants would contribute in reducing near peak and in to some extent peak load and baseload.

Although these sources alone would not be ideal to be used as the main or only source of electricity supply due to their higher capital costs, intermittent nature, requirement for substantial storage and uncertain power supply, they could well achieve significant penetration levels in the Afghan grid especially when all of the islanded grid segments are interconnected and could possibly export power. In the meantime, these sources are environmentally friendly sources of power generation and offer power with pretty stable prices.

In most developed countries, the best way to increase share of intermittent renewables in the existing grid is to transform them into more flexible ones. On the other hand, developing countries like Afghanistan still have the opportunity to think ahead and design power systems that could accommodate higher levels of variable power sources as they are still growing and taking shape. In the meantime, shares of renewables could go

69 higher if the Afghan government sets national targets and designs effective policies. In the end, higher solar PV and wind penetration means less reliance on unpredictable and unstable power purchase agreements with neighboring countries, longer life of limited domestic fossil fuel resources such as coal and natural gas, and less imports of diesel fuel with rising costs and unfriendly environmental impacts.

70

REFERENCES

ACCU. (2010). Retrieved from ACCU Asia-Pacific Cultural Center for UNESCO: http://www.accu.or.jp/esd/forum_esd_2010/program/program12_01/pdf/presentation 5.pdf

ACEP . (2011). Bamiyan 1 MWp PV Power System vs. Topchi 1 MWp MHP . Kabul: Afghanistan Clean Energy Program (ACEP).

ACEP. (2010). Kandahar City Utility Interconnected 10 MW Photovoltaic Power System Feasibility Assessment. Kabul: Afghanistan Clean Energy Program (ACEP).

ADB. (2013). Energy Outlook for Asia and Pacific. Manila: Asian Development Bank (ADB).

AEIC. (2012, February 20). Energy Production. Retrieved from Afghan Energy Information Center (AEIC): http://www.afghaneic.org/production.php

AWS Scientific. (1997, April ). Wind Research: Wind Resource Assessment. Retrieved from National Renewable Energy Laboratoy (NREL): http://www.nrel.gov/wind/pdfs/22223.pdf

AWS Truewind. (2009). Development of Eastern Regional Wind Resource and Wind Plant Output Datasets. Golden, Colorado: National Renewable Energy Laboratory (NREL).

Burton, T., Sharp , D., Jenkins, N., & Bossanyi, E. (2001). Wind Energy Handbook. Chichester: John Wiley & Sons.

CIA. (2014, March 31). The World Fact Book: Afghanistan Economy. Retrieved from Central Intelligence Agency (CIA): https://www.cia.gov/library/publications/the-world- factbook/geos/af.html

CSO. (2014, April 9). Population Statistics. Retrieved from Central Statistics Organization Islamic Republic of Afghanistan: http://cso.gov.af

Ershad, A. (2014). Technical Potential of Solar Photovoltaic Power Plants in Afghanistan. Not Yet Published.

Fichtner. (2013). Islamic Repulic of Afghanistan: Power Sector Master Plan. Stuttgart: Fichtner.

71

GWEC, G. (2014, February). Global Figures. Retrieved from www.gwec.net.

IEA. (2013, November 12). World Energy Outlook. Retrieved from International Energy Agency: http://www.worldenergyoutlook.org/publications/weo-2013/

IPCC. (2007). IPCC Fourth Assessment Report: Climate Change 2007. Retrieved from Intergovernmental Panel on Climate Change (IPCC): http://www.ipcc.ch/publications_and_data/ar4/wg3/en/spmsspm-b.html

Janardan , R., & Nelson, V. (1994). Wind Characteristics. Canyon, Texas: Alternative Energy Institute, West Texas A & M University.

Lambart, T. (2004). Diurnal Pattern Strength. Boulder: Homer Energy .

Lambert, T. (2004). Autocorrelation. Boulder: Homer Energy .

Manwell, J. F., McGowan, J. G., & Rogers, A. L. (2009). Wind Energy Explained. Chichester: John Wiley & Sons Ltd.,.

Messenger, R., & Ventre, J. (2003). Photovoltaic Systems Engineering. Boca Raton, Florida: CRC Press.

MEW . (2013, March 7). Renewable Energy Department. Retrieved from Minstry of Energy and Water Islamic Republic of Afghanistan: http://arbm-mew.gov.af/renewable-energy/

NOAA. (2008, August 20). Climate of Afghanistan. Retrieved from National Climate Data Center: http://www.ncdc.noaa.gov/oa/climate/afghan/afghan-narrative.html

NRECA . (2010). Integrating Wind Generation into the Afghan Transmission Network in Herat and Balkh. Kabul: NRECA.

NREL. (2007, May). International Activities: Afghanistan Resource Maps and Toolkits. Retrieved from National Renewable Energy Laboratory (NREL): http://www.nrel.gov/international/ra_afghanistan.html

Ray, M. L., Rogers, A. L., & McGowan, J. G. (2006). Analysis of Wind Shear Models and and Trends in Different Terrains. Retrieved from University of Massachussets: http://www.umass.edu/windenergy/publications/published/2006/AWEA%202006%20 Wind%20Shear.pdf

REN21. (2013). Renewables 2013 Global Status Report. Paris: REN21 Secretariat.

Seguro, J., & Lambert, T. (2000). Modern estimation of the parameters of the Weibull wind speed distribution for wind energy analysis. Journal of Wind Engineering and Industrial Aerodynamics, 75-84.

72

TETRA TECH. (2009). Development of Wind Energy Meteorology and Engineering for Siting and Design of Wind Energy Projects in Afghanistan. Langhorne: TETRA TECH.

Ueckerdt, F., Brecha, R., & Luderer, G. (2014). Analyzing major challenges of wind and solar variability in power systems. Renewable Energy, Not Yet Published.

UNDP. (2013, March 2013). 2013 Human Development Report. Retrieved from United Nations Development Program Human Development Reports: http://hdr.undp.org/en/content/human-development-report-2013-media-toolkit

Wan, Y.-H., Ela, E., & Orwig, K. (2010). Development of an Equivalent Wind Plant Power-Curve. WindPower 2010. Dallas, Texas: National Renewable Energy Laboratory (NREL).

WB. (n.d.). Afghanistan. Retrieved April 9, 2014, from The World Bank: http://data.worldbank.org/country/afghanistan

73