Electrical design and operation of sustainable business parks
Simon De Clercq
Supervisor: Prof. dr. ir. Lieven Vandevelde Counsellor: Dr. Brecht Zwaenepoel
Master's dissertation submitted in order to obtain the academic degree of Master of Science in Electromechanical Engineering
Department of Electrical Energy, Metals, Mechanical Constructions & Systems Chair: Prof. dr. ir. Luc Dupré Faculty of Engineering and Architecture Academic year 2016-2017
Electrical design and operation of sustainable business parks
Simon De Clercq
Supervisor: Prof. dr. ir. Lieven Vandevelde Counsellor: Dr. Brecht Zwaenepoel
Master's dissertation submitted in order to obtain the academic degree of Master of Science in Electromechanical Engineering
Department of Electrical Energy, Metals, Mechanical Constructions & Systems Chair: Prof. dr. ir. Luc Dupré Faculty of Engineering and Architecture Academic year 2016-2017 The author gives permission to make this master dissertation available for consultation and to copy parts of this master dissertation for personal use. In the case of any other use, the copyright terms have to be respected, in particular with regard to the obligation to state expressly the source when quoting results from this master dissertation.
Ghent, June 2017
The promotor The advisor The author
Prof. dr. ir. L. Vandevelde Dr. B. Zwaenepoel Simon De Clercq Preface
It is a pleasure to be able to finish my engineering studies by writing this dissertation thesis. This work, and the learning process that preceded it, have allowed me to explore the mul- tidisciplinary world of industrial symbiosis and the integration of renewable power sources. It has taught me to enjoy the freedom of independent research and gave me the opportunity to broaden my vision on renewable energy.
I would like to thank my promotors professor Vandevelde and Brecht Zwaenepoel for their support and guidance over the course of this work. I also particularly thank Hendrik Ver- meersch for his input throughout the year; our conversations on the Belgian energy sector were a pleasure. I would like to gratefully acknowledge Christof Deckmyn for his explanation on genetic algorithms.
I want to thank Mieke Gevaerts at SOGent for introducing me to Eiland Zwijnaarde and showing me a glimpse of what urban development is.
To everyone who took their time to review my writing, Brecht, Hendrik and Samie: thank you very much.
Finally, I would like to thank everyone who was there for me, my family, friends, fellow students, house mates, . . . Thank you for making this year so enjoyable. Summary
Electrical design and operation of sustainable business parks Simon De Clercq
Supervisor: Prof. dr. ir. Lieven Vandevelde Counsellor: Dr. Brecht Zwaenepoel
Faculty of Engineering and Architecture Ghent University Academic Year 2016-2017
Abstract - In this study a techno-economic optimisation is carried out on the design and operation of energy systems on newly-developed industrial parks. An optimal sizing model is constructed that uses a genetic algorithm to find the optimal system configuration of an on- grid power system. The system is evaluated according to its life cycle cost and green house gas emissions. While optimal sizing algorithms have been developed for different types of hybrid power systems, these often do not consider the socio-organisational drivers and limitations for inter-firm energy supply facilities. This work presents a list of recommendations for industrial park developers to determine and implement the park’s optimal power system. The current development of the project Eiland Zwijnaarde in Ghent provides the basis for a concrete case study in which the opportunities for an inter-firm power system are identified. The main results suggest that a significant share of renewable generation is part of the optimal configuration under varying mean electricity and gas prices. A medium sized cogeneration unit can compensate the renewables’ intermittent behaviour and lower the thermal energy cost. Large scale electrical storage is found not to be profitable under the used control structure and tariff scheme. A gradual ingress of firms in the park and the subsequent sloped annual energy demand has a negative effect on the fraction of shared facilities in the power system’s optimal configuration.
Keywords - On-grid hybrid power system, Optimal sizing, Eco-industrial Park, Industrial microgrid, Genetic Algorithm, Fuzzy logic Electrical design and operation of sustainable business parks Simon De Clercq Supervisor(s): Prof. dr. ir. Lieven Vandevelde, Dr. Brecht Zwaenepoel Ghent University, Faculty of Engineering and Architecture, Department of Electrical Energy, Metals, Mechanical constructions & Systems Academic year 2016-2017
Abstract— In this study a techno-economic optimisation is carried out ficantly reduce CO2 emissions and mitigate the effects of global on the design and operation of energy systems on newly-developed indus- warming. trial parks. An optimal sizing model is constructed that uses a genetic algorithm to find the optimal system configuration of an on-grid power An industrial park’s power system is characterised by loc- system. The system is evaluated according to its life cycle cost and green alised electric and thermal loads, high energetic consumption house gas emissions. While optimal sizing algorithms have been developed and spatial opportunities for the integration of renewable en- for different types of hybrid power systems, these often do not consider ergy sources. For these reasons, the prevailing paradigm for the socio-organisational drivers and limitations for inter-firm energy sup- ply facilities. This work presents a list of recommendations for industrial electric distribution on an EIP is that of a microgrid. Accord- park developers to determine and implement the park’s optimal power sys- ing to [6], the microgrid concept assumes a cluster of loads tem. The current development of the project Eiland Zwijnaarde in Ghent and microsources operating as a single controllable system that provides the basis for a concrete case study in which the opportunities for an inter-firm power system are identified. The main results suggest that provides both power and heat to the local area. The compon- a significant share of renewable generation is part of the optimal config- ents in this energy cluster function independently in possible uration under varying mean electricity and gas prices. A medium sized interaction with the surrounding macrogrid [7]. This results in cogeneration unit can compensate the renewables’ intermittent behaviour enhanced local control, which can lower the cost of energy dis- and lower the thermal energy cost. Large scale electrical storage is found not to be profitable under the used control structure and tariff scheme. A tribution, aid the integration of renewable sources and thereby gradual ingress of firms in the park and the subsequent sloped annual en- reduce green house gas emissions [8]. Furthermore, the pos- ergy demand have a negative effect on the fraction of shared facilities in the sibility of islanding improves the system security of supply by power system’s optimal configuration. detaching the microgrid’s reliability from that of the macrogrid. Keywords—On-grid hybrid power system, Optimal sizing, Eco-industrial Park, Industrial microgrid, Genetic Algorithm, Fuzzy logic Several methods for the optimal sizing of hybrid power sys- tems have been developed [9]. In many cases the power sys- tem is a stand-alone microgrid and consists of wind generation, I.INTRODUCTION solar PV, a micro source and electrical storage [10] [11] [12] Limitless economic growth, ecological collapse and resource [13] [14]. An industrial park’s high electrical demand, both in scarcity are forcing industry as a whole to rethink its funda- terms of quantity and quality, makes permanent islanding a com- mental principles and resort to more sustainable practices [1]. plicated design choice [15]. Furthermore, dense electrification As defined in the Brundtland Report [2], sustainable develop- in Flanders means that an industrial park can be connected to ment is development that meets the needs of the present without the macrogrid at a relatively low cost. An on-grid power system compromising the ability of future generations to meet their own is therefore considered in this study. needs. In this discourse, industrial symbiosis has emerged as While cogeneration offers high efficiency and flexibility [16], an approach in which traditionally separate industries collabor- only very few papers dealing with the optimal sizing of hybrid ate in order to find synergies that offer competitive advantages. power system take this feature into account [11]. A CHP model This can involve physical exchange of materials, energy, wa- is developed and included in the system model. ter, etc. [3]. Geographic proximity is an important facilitating A genetic algorithm is selected to carry out the optimisa- factor, which is why eco-industrial parks (EIP), with their col- tion process because of its computational efficiency in multi- lective infrastructure and spatial density of firms, have become objective optimisation problems with a high number of optim- an important topic of study in the field of sustainability [1]. ised variables [10] [11]. It has been used extensively in the op- The transition towards a carbon neutral industrial sector is timisation of hybrid power systems [9] [11]. driven by the global phenomenon of climate change. The IPCC II.HYBRIDPOWERSYSTEMMODEL warns that the continued emission of greenhouse gases will cause further warming of global climate patterns and result in The system model simulates the power flows in the microgrid long-lasting changes in all components of the climate system. on an hourly basis during its lifetime according to given load In Europe in 2014, 26% of CO2 equivalent emissions were due profiles, meteorological data and the system parameters. The to the consumption of electricity and gas [4]. Of all electricity system parameters consist of the installed solar power Pinst.solar, consumption in Europe, 36% is consumed by industry [5]. Op- the installed wind power Pinst.wind, the installed CHP power timizing industrial electricity consumption can therefore signi- Pinst.CHP, the battery storage capacity Einst.bat, and the maximum battery power Pinst.bat. The grid topology is shown in Figure 1. and the park’s electrical demand. If the difference is higher than half of the installed power Pinst.CHP, the unit is turned on and fulfils the demand as far as its operating region allows. This region is linearised and electric production is limited between 50% and 100% of the installed capacity [19] [20]. The nom- inal power-to-heat ratio is 1:1 [21] and there is always maximal heat production. The unit’s thermal and electrical efficiencies are variable and depend on the electrical set point, as shown in Table I. The heat demand that is not produced by the CHP in- stallation is generated using an auxiliary gas boiler with a fixed efficiency of 0.80.
Table I: CHP prime mover efficiencies Electric load level ηel ηth ηtot 0.5 0.20 0.57 0.77 0.75 0.23 0.545 0.775 1 0.26 0.52 0.78
D. Battery storage Electrical power can be stored or withdrawn from an elec- trical storage unit. The storage is controlled using a fuzzy logic based control structure that was adapted from [22]. The main aim of the control structure is to charge the battery when the Fig. 1: Microgrid topology electricity price is low and discharge when the electricty price is high. This means that the battery can charge even though the power flow in the grid is negative, or discharge when the power A. Wind power flow is positive. However, under no circumstances should the Wind power can be modelled using wind speed data and a installed battery require fortification of the grid connection (and turbine’s power curve [10], [17] [18]. However, wind speed data thus a greater grid dependency). i 1 for the region of Flanders is hard to come by or simply does not The input parameters are the previous state of charge SOC − i exist. An alternative to wind speed data is historical data of wind and the instantaneous power flow at time i, PF . These are nor- power production [13] which in Belgium is easily accessible. malised and fuzzified according to 5 categories ranging between Data for on-shore wind generation in Belgium in 2015 is used very low and very high. The linguistic control rules are shown in to calculate the wind power production at every time step. The Table II. The resulting value Pfuzz is withdrawn from the battery normalised hourly time series of capacity factors is multiplied as far as the physical constraints allow. by the installed wind power Pinst.wind to obtain the amount of hourly wind production. Table II: Fuzzy control rules for battery storage i i i price PF Pfuzz i i P = capacity factor Pinst.wind (1) L / VP wind wind · M / Z The annual average capacity factor is equal to 0.244. H / VN The total generated electricity during one year is equal to the VH / VN sum of wind generation during every time step. (L) VH N ¬ (VH) VL P 8760 ¬ i (H) VL P Ewind = Pwind (2) ¬ VL: very low, L: low, M: medium, H: high, VH: very high i=1 X VN: very neg., N: neg, Z: zero, P: pos. VP: very pos. B. Solar power Historical solar production data for the region of East- Four constraints are considered. Flanders is used to calculate the hourly solar generation. The The power constraint dictates that power drawn from or fed annual average power factor is 0.119. to the battery during one time step can not exceed the maximum installed power Pinst.bat [23]. i i P = capacity factor Pinst.solar (3) The efficiency constraint takes into account the power loss solar solar · that occurs when energy is converted in the storage system. In C. Cogeneration theory the charging efficiency ηcharge and the discharging effi- The cogeneration unit follows the electric load according to ciency ηdischarge will have different values. However, separate the instantaneous difference between the renewable production values for these parameters are hard to measure [12] and battery manufacturers will therefore define a round trip efficiency for The initial investment cost depends on the size of the installed one entire charge and discharge cycle. In the model used here power and is equal to the sum of the investment cost for the sep- the charge and discharge efficiency are both assumed to be equal arate components according to the parameters in Table III. The to the square root of the round trip efficiency. ηbat = ηcharge = batteries have a double investment cost because of their smaller ηdischarge = √ηround = 0.90. life time. Storage systems exhibit self-discharge behaviour, which Cinitial = CAPEXsolar Pinst.solar + CAPEXwind Pinst.wind means that over time a certain fraction of the stored power will · · +2 CAPEX E + CAPEX P (9) be dissipated and lost [24] [17] [15]. The rate of self-discharge · c,bat · inst.bat p,bat · inst.bat per time step is denoted by the symbol σ. +CAPEX P CHP · inst.CHP By introducing a maximum depth of discharge of the elec- The annual cost is equal to the sum of the operational costs ac- trical storage unit, the longevity of the system can be prolonged cording to Table III. [24] [25]. Concretely this means that when the battery is at its minimum state of charge, it no longer allows power to be Cj = OPEX E + OPEX E annual solar · solar wind · wind withdrawn. SOCmax is in practice almost always equal to 1 and +OPEXf,CHP + OPEXv,CHP ECHP (10) SOCmin is here equal to 0.15. · The four constraints can mathematically be expressed as +grid cost + fuel cost shown below. i Electricity is bought (Pgrid < 0) at the instantaneous Belpex 0 < P exi < P (4) | bat| bat price plus an additional fee for the network costs. During time i i 1 i SOC = (1 σ) SOC − (5) steps with surplus electricity (Pgrid > 0) it is sold at the Belpex − · i price. The annual grid cost is calculated according to Equa- i 1 P ex i SOC − + η bat P ex < 0 tion 11. i bat Ebat bat SOC = · i (6) i 1 1 P ex i (SOC − + bat P ex > 0 8760 ηbat · Ebat bat i i i i (Pgrid < 0) (Belpex + fee) + (Pgrid > 0) Belpex (11) i · · SOC < SOC < SOC (7) i=1 min max X E. Demand Response The fuel consumption at every time step is calculated using the CHP set point, the CHP efficiency, and the auxiliary boiler’s In this study a heuristic approach is used to model demand consumption. The total annual fuel cost is the sum of the con- response. It looks at the daily interaction with the utility grid sumption multiplied by the gas price. and shifts a certain percentage of the total daily energy profile 8760 from times of high to times of low consumption. An example fuel consumptioni gas price (12) for the original and resulting power flow for one week is shown · i=1 in Figure 2. Using this method, a shift of 2.5% of the daily X energy interaction is found to correspond to an annual power peak reduction by 10.7%. Table III: Economic parameters [27] [28] Battery Solar Wind CHP Capacity Power 500 without DR CAPEX (e/kW) 1570 1911 3940 175 175 with DR 0 OPEXv (e/kWh) 0.021 0.017 0.0108 - - OPEXf (e/year) - - 9345 - - -500 Lifetime (year) 25 25 25 12.5
-1000
Grid power in kW -1500 B. Emissions evaluation
-2000 The microgrid’s annual CO2 emissions are calculated ac- cording to the total annual production per source and their -2500 0 20 40 60 80 100 120 140 160 technology-specific reference values (Table IV). hour tot rel.solar rel.wind Fig. 2: Demand response model CO2 = CO2 Esolar + CO2 Ewind+ · · (13) rel.CHP rel.grid CO2 ECHP + CO2 Egrid III.SYSTEM EVALUATION AND OPTIMISATION · · In the case of a netto export of electricity to the grid, the emis- A. Economic evaluation sions due to the additional electricity production are subtracted The system is evaluated according to its life cycle cost (LCC). from the microgrid’s emissions. The relative share of each of This consists of the initial investment cost Cinitial and the dis- the microgrid’s generation units allows a relative value for the j counted sum of the annual cost Cannual for each year [10] [26] CO2 emissions per kWh of produced power to be calculated. [15]. rel.solar rel.wind rel.CHP LT j CO E + CO E + CO E C COmix = 2 · solar 2 · wind 2 · CHP LCC = C + annual (8) 2 E + E + E initial (1 + r)j solar wind CHP j=1 (14) X rel.grid For netto export, this value replace the value of CO2 in Figure 5 shows the microgrid’s hourly production profiles for Equation 13. one week in the summer. There are strong daily generation peaks due to the solar power production. These are almost dir- Table IV: Relative CO2 emissions [29] ectly translated into energy export peaks towards the grid. The Solar Wind CHP Grid CHP mainly works at night to compensate the absence of solar rel CO2 (gCO2/kWh) 41 11 300 210 generation. There is a very clear netto positive power exchange with the grid. The value for the LCC in this scenario is 188 million e. IV. EILAND ZWIJNAARDE 100.24 gCO2 is emitted per consumed kWh. The LCC for a Eiland Zwijnaarde is an industrial park in the region of Ghent reference case with the same electricity and gas buying price e that is currently under development. The developers are assess- but without on-site production is 243 million . The reference ing the possibilities of collective energy infrastructure and it is emissions are equal to the grids emissions, 210 gCO2/kWh. The therefore an appropriate case study for the developed model. proposed power system thus offers an improvement according The site’s annual electric consumption is estimated around 60 to both evaluation methods.
GWh. and the annual thermal demand at 40 GWh. The electric ×104 load profile is based on the consumption profile of a technology 1.5 park in the region of Ghent (Figure 3). The thermal load pro- file is a generic load profile from the Flemish Energy Regulator 1 (Figure 4). The peak electrical demand is 12 MW and the peak 0.5 thermal demand is 14 MW.
0
12000 15000 Average power (kW) 11000 -0.5 P solar 10000 P wind 10000 9000 -1 P CHP P 8000 grid Load profile 7000 -1.5 5000 20 40 60 80 100 120 140 160 Electric load profile (kW) 6000 Thermal load profile (kW) Time (hours) 5000 Fig. 5: Hourly production profile over one week 4000 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 hour hour Fig. 3: Electric load profile Fig. 4: Thermal load profile A. DR Demand response is included in the model as described The upper boundary for the optimal size of the installed tech- above. On a daily basis 2.5% of the total electrical demand is nologies are derived from the physical boundaries of the site. In shifted from times of high demand to times of low demand. total there is 350000 hectares available for companies. At a con- The resulting optimal installed power for each technology is 2 shown in Figure 6. There is a general decrease in installed power servative estimate of 10m /kWinst. for solar power, this means that the maximum amount of solar peak power that can be in- compared to the optimisation results without demand response. stalled is 35MW. For wind power it is estimated that there is The total size reduction is about 15%. The reduction is most space to install 3 large turbines. At 3.3 MW/turbine, the res- pronounced for solar power and the grid connection; both de- ulting upper boundary for installed peak power is 10 MW. The crease by more than 3MW. In terms of energy, shown in Figure upper boundary for CHP is the electric peak demand, 10 MW. 7, there is less solar power production and a subsequently smal- There is no upper boundary for the installed battery power or its ler energy export to the utility grid. Overall, there is a smaller energetic capacity. interaction with the utility grid while maintaining a high share of renewable power. V. CASE STUDY AND SIMULATION RESULTS The optimal sizing results for an average electricity buying 30 DR No DR price of 115 e/MWh and a gas price of 63 e/MWh are tabulated 20 below. Initially, demand response is not considered.
Psolar : 24.70 MW Esolar : 25700 MWh/year (MW) 10 Pwind : 9.91 MW Ewind : 21150 MWh/year Installed power 0 PCHP : 3.56 MW ECHP : 17070 MWh/year Solar Wind CHP Grid Pgrid : 17.40 MW Egrid : -3925 MWh/year Fig. 6: Installed power with demand response Ebat : 18 kWh There is a high share of both solar and wind power in the total annual electricity generation. The amount of wind power is just below the site’s upper boundary and there is a netto export of B. Gradual ingress electricity to the utility grid. The amount of installed battery In the previous sections the hybrid power system was sized power is negligible on this scale. according to an equal energetic demand for every year of its life- 104 104 · · 3 DR No DR 3 Ingress No ingress 2 2 1 1
(MWh/year) 0 (MWh/year) 0 Generated power Generated power -1 -1 Solar Wind CHP Grid Solar Wind CHP Grid Fig. 7: Generated power with demand reponse Fig. 9: Generated power with gradual ingress time. However, a gradual ingress of companies on the site will rapidly for an increasing value of LCC. On average, a reduc- result in a much smaller load profile during the first years of the tion of 1 gCO2/kWh comes at an additional cost of 117000 e park. In the model, the first year’s electric and thermal demand over the lifetime of the installation. This corresponds to 4680 e are set to 10% of the nominal value. During the following ten on an annual basis. For an LCC greater than 19 million e the years the demands increase by 10% every year. After 10 years curve is practically linear and the total cost of saving an ad- the electric and thermal demand are at their nominal value. The ditional gC02/kWh is equal to 340000 e, or an annual cost of power system is sized once and the installed capacities are fixed 13600 e/year . during the entire lifetime of the installation.
The simulation results are shown in Figures 8 and 9. There is Pareto front a clear absence of cogeneration in the optimised system. In- 100 stead, all of the thermal energy is provided by the auxiliary 95 boiler. The CHP unit is controlled according to the electrical 90 demand, which means that during the first years it will prac- 85 tically never be turned on. Instead the electrical power mainly 80 comes from the grid and the wind turbines. While in the previ- 75 ous optimisations there was always a netto export of electricity 70 to the grid, in this case there is a large netto import. This grid 65 dependency does not result in a larger grid connection; the grid’s GHG emissions in gCO2/kWh 60 installed power decreases from around 18 MW to 13 MW. This 55 is due to the lower installed solar fraction. Solar power gives 50 1.88 1.9 1.92 1.94 1.96 1.98 2 2.02 2.04 large production peaks during summer which requires a strong Stop Pause ×108 grid connection to inject the energy surplus on the grid. De- Fig. 10: Pareto curve creasing the solar fraction thus means that the grid connection can be made smaller. Figure 11 shows each technology’s annually generated elec- The relative CO2 emissions are 95.7 gCO2/kWh. This is tricity for every point on the pareto front. For low LCC, the slightly lower than the reference scenario without gradual in- reduction in CO2 is due to a decrease in cogeneration and an gress. increase in solar power. This corresponds to the steepest part of the pareto curve, which means that this reduction comes at 30 the lowest cost per gCO2 saved. The wind power fraction and Ingress No ingress grid export stay relatively constant. For increasing LCC, the 20 generated solar power reaches its maximum. For a further CO2
(MW) 10 reduction the installed power of the CHP keeps decreasing and the energy deficit is compensated by less electricity exports to Installed power 0 the grid. This way the emissions can be reduced down to about Solar Wind CHP Grid 54 gCO2/kWh. In comparison, the average in electricity in Bel- Fig. 8: Installed power with gradual ingress gium are between 180 and 210 gCO2/kWh [30] and in Europe this is around 320 gCO2/kWh [31]. For very low values of CO2 the netto power exchange with the grid becomes positive. C. Multi-dimensional optimisation VI.CONCLUSION A multi-objective optimisation is carried out that finds the nondominated solution set for the LCC and the CO2 emissions The optimal system is assessed for a range of scenarios. The per kWh. The pareto curve is shown in Figure 10. results suggest a large scale roll-out of renewable generation in The configuration with the lowest LCC and the highest CO2 practically all cases. For maximal installed renewable produc- lies around 18.8 million e. This corresponds to the result of the tion the renewable power sources are theoretically able to supply single-objective optimisation. The CO2 emissions initially drop all the park’s annual energy demand. However, the intermittent ×107 3.5 [9] S. R. Tito, T. T. Lie, and T. N. Anderson, “Optimal sizing of a wind- photovoltaic-battery hybrid renewable energy system considering socio- 3 demographic factors,” Solar Energy, vol. 136, pp. 525–532, 2016. [10] E. Koutroulis, D. Kolokotsa, A. Potirakis, and K. Kalaitzakis, “Method- 2.5 ology for optimal sizing of stand-alone photovoltaic/wind-generator sys- 2 tems using genetic algorithms,” Solar Energy, vol. 80, no. 9, pp. 1072– E solar 1088, 2006. 1.5 E wind [11] M. S. Ismail, M. Moghavvemi, and T. M. I. Mahlia, “Genetic algorithm E CHP based optimization on modeling and design of hybrid renewable energy 1 E grid systems,” Energy Conversion and Management, vol. 85, pp. 120–130, 0.5 2014. Generated power (kWh) [12] H. Yang, Z. Wei, and L. Chengzhi, “Optimal design and techno-economic 0 analysis of a hybrid solar-wind power generation system,” Applied Energy, -0.5 vol. 86, no. 2, pp. 163–169, 2009. [13] B. Zhao, X. Zhang, P. Li, K. Wang, M. Xue, and C. Wang, “Optimal sizing, -1 operating strategy and operational experience of a stand-alone microgrid 95.6gCO /kWh 71.74gCO /kWh 53.94gCO /kWh 2 2 2 on Dongfushan Island,” Applied Energy, vol. 113, pp. 1656–1666, 2014. [14] R. Dufo-Lopez´ and J. L. Bernal-Agust´ın, “Multi-objective design of Fig. 11: Generated power on pareto front PV-wind-diesel-hydrogen-battery systems,” Renewable Energy, vol. 33, no. 12, pp. 2559–2572, 2008. [15] W. L. Theo, J. S. Lim, S. R. Wan Alwi, N. E. Mohammad Rozali, W. S. Ho, and Z. Abdul-Manan, “An MILP model for cost-optimal planning of behaviour of these sources and the current price structure means an on-grid hybrid power system for an eco-industrial park,” Energy, vol. that a solid grid connection is in generally favourable. 116, pp. 1423–1441, 2016. [16] B. Zwaenepoel, J. Vansteenbrugge, T. L. Vandoorn, G. Van Eetvelde, and A small amount of demand response can have a large impact L. Vandevelde, “Renewable energy balancing with thermal grid support,” on the size of the installation. A power system with DR is more Chemical Engineering Transactions, vol. 35, pp. 535–540, 2013. self-sufficient without increasing the fraction of non-renewable [17] Y. Li, X. Chen, J. Xiao, and X. Wei, “Optimal configuration for distributed generations in micro-grid system considering diesel as the main control production. source,” in International Conference on Advanced Mechatronic Systems, A gradual ingress of firms in the park and the subsequent 2014, pp. 552–556. [18] S. M. M. Tafreshi, H. A. Zamani, S. M. Ezzati, M. Baghdadi, and H. Va- sloped annual energy demand has a very significant effect on hedi, “Optimal unit sizing of distributed energy resources in microgrid the power system’s optimal configuration. Dynamically sizing using genetic algorithm,” Proceedings - 2010 18th Iranian Conference on the site’s power system according to a changing annual energy Electrical Engineering, ICEE 2010, pp. 836–841, 2010. [19] H. Wang, W. Yin, E. Abdollahi, R. Lahdelma, and W. Jiao, “Modelling and demand could be a future topic of research to address this issue. optimization of CHP based district heating system with renewable energy The implemented storage control structure is not able to offer production and energy storage,” Applied Energy, vol. 159, pp. 401–421, 2015. any economic incentives for the large scale deployment of bat- [20] M. Nazari-Heris, S. Abapour, and B. Mohammadi-Ivatloo, “Optimal eco- tery storage. This suggests that new remuneration methods are nomic dispatch of FC-CHP based heat and power micro-grids,” Applied needed in order to successfully introduce this technology. Thermal Engineering, vol. 114, pp. 756–769, 2017. [21] P. Ghadimi, S. Kara, and B. Kornfeld, “The optimal selection of on-site A multi-objective optimisation identifies the specific cost of CHP systems through integrated sizing and operational strategy,” Applied decreasing CO2 emissions. Increasing the solar share and redu- Energy, vol. 126, pp. 38–46, 2014. [22] J. P. Fossati, A. Galarza, A. Martin-Villate, J. M. Echeverria, and cing the size of the CHP installation results in reduced emissions L. Fontan, “Optimal scheduling of a microgrid with a fuzzy logic con- but comes at an additional cost. trolled storage system,” International Journal of Electrical Power and En- While in all scenarios there is a high share of renewable gen- ergy Systems, vol. 68, pp. 61–70, 2015. [23] J. P. Fossati, A. Galarza, A. Martin-Villate, and L. Fontan, “A method for eration, there is limit to the physical amount of energy that can optimal sizing energy storage systems for microgrids,” Renewable Energy, be generated on-site. Solar and wind power clearly play an im- vol. 77, pp. 539–549, 2015. portant role in the optimal power system, but their share is lim- [24] H. Yang, W. Zhou, L. Lu, and Z. Fang, “Optimal sizing method for stand- alone hybrid solar-wind system with LPSP technology by using genetic ited by the site’s physical boundaries. Demand reduction should algorithm,” Solar Energy, vol. 82, no. 4, pp. 354–367, 2008. therefore be considered in order to make the park independent [25] A. Chaouachi, R. M. Kamel, R. Andoulsi, and K. Nagasaka, “Multiobject- ive Intelligent Energy Management for a Microgrid,” IEEE Transactions of exhaustive energy sources. on Industrial Electronics, vol. 60, no. 4, pp. 1688–1699, 2013. [26] A. Gonzalez,´ J. R. Riba, A. Rius, and R. Puig, “Optimal sizing of a hy- REFERENCES brid grid-connected photovoltaic and wind power system,” Applied En- ergy, vol. 154, pp. 752–762, 2015. [1] H. Zhao, H. Zhao, and S. Guo, “Evaluating the comprehensive benefit [27] International Energy Agency, Nuclear Energy Agency, and Organisation of eco-industrial parks by employing multi-criteria decision making ap- for Economic Co-operation and Development, “Projected Costs of Gener- proach for circular economy,” Journal of Cleaner Production, vol. 142, ating Electricity 2015,” Tech. Rep., 2015. pp. 2262–2276, 2017. [28] K. Darrow, R. Tidball, J. Wang, and A. Hampson, “Catolog of CHP [2] United Nations World Commission on Environment and Development, Technologies,” U.S. Environmental Protection Agency, Tech. Rep. March, “Our Common Future,” 1987. 2015. [3] M. R. Chertow, “Industrial Symbiosis: Literature and Taxonomy,” Annual [29] IPCC, “Climate Change 2013 - The Physical Science Basis,” Tech. Rep., Review of Energy and the Environment, vol. 25, no. 1, pp. 313–337, 2000. 2014. [4] Eurostat, “Energy balance sheets,” pp. 345–52, 2016. [30] M. Messagie, J. Mertens, L. Oliveira, S. Rangaraju, J. Sanfelix, T. Coose- [5] European Environmental Agency, “Final electricity consumption by sector mans, J. Van Mierlo, and C. Macharis, “The hourly life cycle carbon foot- , EU-27,” p. European Environment Agency, 2013. print of electricity generation in Belgium, bringing a temporal resolution [6] R. Lasseter, “MicroGrids,” in IEEE Power Engineering Society Winter in life cycle assessment,” Applied Energy, vol. 134, no. August, pp. 469– Meeting. Conference Proceedings, vol. 1, 2002, pp. 305–308. 476, 2014. [7] K. V. Hende and C. Wouters, “The Regulation of Microgrids in Liberalized [31] F. Millo, L. Rolando, R. Fuso, and F. Mallamo, “Real CO2 emissions be- Electricity Markets in the EU and East Asia,” European Networks Law and nefits and end user’s operating costs of a plug-in Hybrid Electric Vehicle,” Regulation Quarterly, vol. 3, pp. 190–204, 2014. Applied Energy, vol. 114, pp. 563–571, 2014. [8] R. Hanna, M. Ghonima, J. Kleissl, G. Tynan, and D. G. Victor, “Evaluating business models for microgrids: Interactions of technology and policy,” Energy Policy, vol. 103, no. August 2016, pp. 47–61, 2017. Contents
1 Introduction 1
1.1 Eco-industrial parks ...... 2
1.1.1 Definitions and concepts ...... 2
1.1.2 Drivers and limitations of eco-industrial parks ...... 3
1.1.3 Inter-firm energy supply ...... 5
1.1.4 Industrial microgrids ...... 6
1.2 Eiland Zwijnaarde ...... 9
1.2.1 Eiland Zwijnaarde as an eco-industrial park ...... 10
1.2.2 Vision of project ...... 10
1.2.3 Current state of affairs ...... 13
2 Microgrids in the Belgian electricity sector 15
2.1 The liberalised electricity market ...... 15
2.2 Price of electricity ...... 18
2.2.1 Commodity price ...... 19
2.2.2 Transmission tarrifs ...... 20
2.2.3 Distribution tarrifs ...... 21
2.2.4 Industrial energy contract ...... 21
2.3 Energy policy in Flanders ...... 22
2.4 CO2 market ...... 24
2.5 Grid structure and microgrid opportunities ...... 25
2.5.1 Functionality and economic incentives ...... 25
xiii 2.5.2 Control ...... 27
2.5.3 Ownership ...... 27
3 Hybrid Power systems 30
3.1 Wind power ...... 31
3.1.1 Wind turbine power curve ...... 32
3.1.2 Wind speed data ...... 33
3.1.3 Dynamic economic analysis ...... 35
3.2 Solar power ...... 37
3.2.1 PV-curve & Maximum power point (MPP) ...... 38
3.2.2 Dynamic economic analysis ...... 40
3.3 Cogeneration ...... 42
3.3.1 Cogeneration technologies ...... 43
3.3.2 Modelling and sizing ...... 45
3.3.3 Modes of operation ...... 50
3.3.4 Dynamic techno-economic analysis ...... 51
3.3.5 System evaluation ...... 53
3.3.6 Reference case ...... 54
3.3.7 Electric load following: ...... 54
3.3.8 Thermal load following: ...... 56
3.3.9 Conclusion on CHP model ...... 58
3.4 Electrical storage ...... 58
3.4.1 Modelling ...... 59
3.4.2 Battery control ...... 62 3.5 Demand-side management ...... 68
3.5.1 Demand response modeling ...... 69
4 Optimal design and sizing 71
4.1 Evaluation of electricity production system ...... 71
4.1.1 Economic evaluation ...... 71
4.1.2 Technical evaluation and constraints ...... 75
4.2 Optimised variables ...... 77
4.3 Multi-objective optimisation ...... 78
4.3.1 Pareto front ...... 78
4.4 Genetic Algorithm ...... 79
5 Simulation results and discussion 82
5.1 Energy key figures and model parameters ...... 82
5.2 Sensitivity analysis ...... 83
5.2.1 System evaluation ...... 84
5.2.2 Installed power per technology ...... 85
5.2.3 Average weekly power flow ...... 87
5.3 Future prices ...... 90
5.4 Impact of ETS market ...... 93
5.5 Impact of demand response ...... 94
5.6 Impact of gradual ingress ...... 96
5.7 Multi-dimensional analysis ...... 100
6 Conclusion 103
List of Abbreviations and Symbols
ARP Access responsible party ACS Annualized cost of system BREAAM Building Research Establishment Environmental Assessment Method
Cinitial Initial investment cost
Cannual Annual system cost
CAPEXsolar Capital expenditure for solar power
CAPEXwind Capital expenditure for wind power
CAPEXCHP Capital expenditure for CHP unit
CAPEXc,bat Capital expenditure for battery capacity
CAPEXp,bat Capital expenditure for battery power CHP Combined heat and power
CO2,mix Average CO2 emissions per kWh of electricity produced locally in the microgrid COE Cost of electricity DR Demand response DSM Demand side management DSO Distribution system operator
ECHP Annually produced electrical CHP power
Egrid Annual power exchange with macrogrid
Esolar Annually produced solar power
Ewind Annually produced wind power
Einst.bat Maximum amount of energy that can be stored in the storage system EE Embodied energy EES Electrical energy storage EIP Eco-industrial park ESCO Energy service company ETS Emission trading system GHG Green house gas LCC Life cycle cost i LPE Instantaneous electrical demand i LPT Instantaneous thermal demand LPSP Loss of power supply probability LT Lifetime MPP PV maximum power point MOC Mean operation cost
OPEXf,CHP Fixed operational expenditure CHP
OPEXsolar Operational expenditure solar power
OPEXv,CHP Variable operational expenditure CHP
OPEXwind Operational expenditure wind power OTC Over the counter PB Power balance i P exbat Instantaneous power exchange with storage system PF i Instantaneous difference between the electric production and the electric consumption i PFnorm Normalised instantaneous power flow Pgrid1 Grid interaction because of the battery’s power limitation
Pgrid2 Grid interaction because of the battery’s state of charge limitation
PI/O Reference value for CHP on-off decision i PCHP Instantaneous CHP electrical power production i Pfuzz Battery’s instantaneous economic set point i Pgrid Instantaneous interaction with macrogrid i Psolar Instantaneous solar power production i Pwind Instantaneous wind power production Pinst.bat Maximum amount of power that can be drawn or stored in the battery during one time step
Pinst.CHP Installed electrical CHP power
Pinst.grid Installed grid power
Pinst.solar Installed solar power
Pinst.wind Installed wind power PCC Point of common coupling P-Q ratio Power to heat ratio i pricenorm Normalised instantaneous electricity price PV Photovoltaic i QCHP Instantaneous CHP heat production i Qboiler Instantaneous boiler heat production r Interest rate REP Renewable energy penetration ROI Return on investment SOCi State of charge
SOCmin EES minimum state of charge
SOCmax EES maximum state of charge SPV Special purpose vehicle TSO Transmission system operator UPS Uninterruptible power supply vi Instantaneous wind speed vci Cut-in wind speed vr Rated wind speed vco Cut-out wind speed σ EES hourly self-discharge coefficient
ηbat EES charging and discharging efficiency
ηel Electrical CHP efficiency
ηth Thermal CHP efficiency
ηtot Total CHP efficiency List of Figures
1.1 Classification of industrial parks ...... 3
1.2 Microgrid topology ...... 8
1.3 Map of the site Eiland Zwijnaarde ...... 9
1.4 Development scenarios for Eiland Zwijnaarde ...... 11
1.5 Trias Energetica ...... 13
2.1 Lay-out of liberalised electricity market ...... 16
2.2 Breakdown of the price of electricity ...... 19
2.3 Hourly Belpex price in 2015 ...... 19
2.4 Projection of ETS CO2 price ...... 25
3.1 Wind turbine power curve ...... 32
3.2 Wind speed profiles based on the log and power law ...... 34
3.3 Historical wind production data ...... 35
3.4 Hourly load profile ...... 37
3.5 Hourly wind power production ...... 37
3.6 Hourly power exchange with the grid ...... 37
3.7 Power flows over one week ...... 37
3.8 3D Simulation results for wind power ...... 38
3.9 Simulation results for specific grid price ...... 39
3.10 PV power-voltage characteristic ...... 40
3.11 Historical solar production data ...... 40
3.12 Hourly solar power production ...... 41
xix 3.13 Power flows over one week ...... 41
3.14 3D Simulation results solar power ...... 42
3.15 Simulation results for specific grid price ...... 42
3.16 Lay-out of a CHP system ...... 43
3.17 CHP operating region ...... 48
3.18 Operational cost of CHP ...... 49
3.19 Linearised CHP operating region ...... 50
3.20 Hourly electric load profile ...... 52
3.21 Hourly thermal load profile ...... 52
3.22 Hourly Belpex price ...... 52
3.23 Electric load following control algorithm ...... 55
3.24 Simulation results with electric load following ...... 56
3.25 Thermal load following control algorithm ...... 57
3.26 Simulation results with thermal load following ...... 57
3.27 Battery control algorithm - maximal usage ...... 64
3.28 Battery control algorithm - fuzzy logic ...... 65
3.29 Membership functions for power flow ...... 67
3.30 Membership functions for electricity price ...... 67
3.31 Membership functions for Pfuzz ...... 68
3.32 Demand response over one day ...... 70
3.33 Demand response over one week ...... 70
3.34 Pgrid without DR (entire year) ...... 70
3.35 Pgrid with DR (entire year) ...... 70 4.1 Graphical representation of a pareto front ...... 79
4.2 Flowchart of a genetic algorithm ...... 80
5.1 Hourly electric load profile ...... 83
5.2 Hourly thermal load profile ...... 83
5.3 Life cycle cost ...... 85
5.4 Cost of electricity ...... 85
5.5 CO2 emissions ...... 86
5.6 Power exchange with grid ...... 86
5.7 Installed solar power ...... 87
5.8 Installed wind power ...... 87
5.9 Installed CHP power ...... 87
5.10 Installed storage capacity ...... 87
5.11 Weekly power flow during the entire year ...... 88
5.12 Hourly simulation during one week in winter ...... 89
5.13 Hourly simulation during one week in summer ...... 90
5.14 Life cycle cost ...... 91
5.15 Cost of electricity ...... 91
5.16 CO2 emissions ...... 91
5.17 Installed solar power ...... 92
5.18 Installed wind power ...... 92
5.19 Installed CHP power ...... 92
5.20 Installed batter capacity ...... 92
5.21 Installed power with ETS ...... 94 5.22 Generated power with ETS ...... 94
5.23 Installed power with DR ...... 96
5.24 Generated power with DR ...... 96
5.25 Effect of ingress on energetic demand ...... 97
5.26 Installed power with ingress ...... 98
5.27 Generated power with ingress ...... 98
5.28 Weekly power flow for the fourth and tenth year ...... 99
5.29 Hourly power curves over one week in summer for different years ...... 100
5.30 2-dimensional pareto front ...... 101
5.31 Installed power on pareto front ...... 101
5.32 Generated power on pareto front ...... 101 List of Tables
2.1 Example of an industrial energy contract ...... 22
3.1 Overview of technologies in studied literature ...... 31
3.2 Commercial wind turbines ...... 33
3.3 Simulation parameters for wind power ...... 36
3.4 Overview simulation parameters for solar power ...... 41
3.5 Overview of CHP technologies ...... 44
3.6 CHP cost function coefficients ...... 49
3.7 Cogeneration prime mover efficiencies ...... 50
3.8 Overview simulation parameters for cogeneration case ...... 53
3.9 Technical parameters for different storage technologies ...... 59
3.10 Economic parameters for storage technologies ...... 60
3.11 Fuzzy control rules for battery storage ...... 68
3.12 Demand response simulation results ...... 70
4.1 Overview of optimisation methods in studied literature ...... 72
4.2 List of abbreviations and occurrence ...... 73
4.3 CO2 emissions per technology in gCO2eq/kWh ...... 76
5.1 Predicted energetic consumption for different scenarios ...... 82
5.2 Economic simulation parameters ...... 84
5.3 Future economic simulation parameters ...... 90
5.4 Simulation results with ETS ...... 93
5.5 Simulation results with DR ...... 95
xxiii xxiv
5.6 Simulation results with gradual ingress ...... 97
1 Sensitivity analysis for current market prices ...... 119
2 Sensitivity analysis for future market prices ...... 120 Chapter 1
Introduction
Limitless economic growth, ecological collapse and resource scarcity are forcing industry as a whole to rethink its fundamental principles and resort to more sustainable practices [1]. As defined in the Brundtland Report [2], sustainable development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs. In this discourse, industrial symbiosis has emerged as an approach in which traditionally separate industries collaborate in order to find synergies that offer competitive advantages. This can involve physical exchange of materials, energy, water, etc. [3]. Geo- graphic proximity is an important facilitating factor, which is why eco-industrial parks (EIP), with their collective infrastructure and spatial density of firms, have become an important topic of study in the field of sustainability [1].
The transition towards a carbon neutral industrial sector is driven by the global phenomenon of climate change. The IPCC warns that the continued emission of greenhouse gases will cause further warming of global climate patterns and result in long-lasting changes in all components of the climate system. According to their 2014 assessment report, these change will increase the likelihood of severe, pervasive and irreversible impacts for people and ecosystems [4]. In
Europe in 2014, 26% of CO2 equivalent emissions were due to the consumption of electricity and gas [5]. Of all electricity consumption in Europe, 36% is consumed by industry [6].
Optimizing industrial electricity consumption can therefore significantly reduce CO2 emissions and mitigate the effects of global warming.
The aim of this study is to carry out a techno-economic analysis on the design and operation of energy systems on newly-developed industrial parks. A microgrid model is constructed that optimises the system configuration according to its life cycle cost and green house gas emissions. The socio-organisational framework is assessed by means of a literature study which results in a list of recommendations for park developers. The current development of the project Eiland Zwijnaarde in Ghent provides the basis for a concrete case study in which the opportunities for an inter-firm power system are identified.
1 Chapter 1. Introduction 2
1.1 Eco-industrial parks
1.1.1 Definitions and concepts
What sets EIPs apart from conventional business parks is that the ecological footprint and carbon dioxide emissions of the individual businesses as well as the park as a whole are actively reduced. Energy and construction technologies are employed to optimise energy efficiency and material cycles. An integrated management system facilitates inter-firm exchange of energy, materials and information. A regulatory system encourages the companies to meet their performance goals [7].
Four main definitions in the field of eco-industrial parks can be identified. They are schem- atically represented in Figure 1.1.
Sustainable industrial parks A sustainable industrial park is the most far-reaching concept in terms of sustainability and collectivity in the context of industrial park design. Technological, economic, social and spatial opportunities offered by the geographical proximity are combined in ways that go beyond the firms’ individuality [8]. This results in a collective and integrated approach regarding the supply and exchange of energy and materials, use of equipment and facilities, etc. Sustainable industrial parks are well-integrated in the surrounding area and require extensive measures regarding ownership and stakeholders relationships.
Eco-industrial parks In EIPs, individual companies exploit inter-firm synergies in term of materials, energy and water. There is a strong focus on industrial symbiosis between the participants, although the firms pursue their individual objectives. An eco-industrial park can be seen as the composition of a number of industrial symbiosis projects between different firms in the same location [9]. It forms an industrial ecosystem in which the consumption of energy and materials is optimised, waste generations is minimised and residues, whether in terms of materials or energy, are optimally reinjected in the value chain [10].
Green industrial parks A green industrial park is a collection of firms that individually strive towards sustain- ability. Possibilities to exploit inter-firm synergies are not actively investigated. The majority of sites in Flanders with an eco-label belong to this group [11].
Low carbon industrial parks Low carbon industrial parks is the combination of green industrial parks and eco- industrial parks. Firms focus on renewable energy and clean processes on an individual Chapter 1. Introduction 3
level while looking for opportunities of interaction with neighbouring sites.
Sustainable Eco-industrial Green industrial parks parks industrial parks
Low carbon industrial parks
Figure 1.1: Classification of industrial parks [11]
1.1.2 Drivers and limitations of eco-industrial parks
Since the concept of industrial symbiosis was first used in academic literature in the 1980’s, there have been many attempts to harness its advantages [12]. A review of literature is carried out to identify the drivers and limitations of the implementation of industrial ecology concepts in industrial parks. Many of the cases cover a broad range of technologies, from energy and resource supply to transport, employee facilities and water treatment. The scope of this study is confined to the possibilities regarding collective supply of electrical power. The fact that eco-industrial parks are such multi-disciplinary entities makes them hard to study. The technical aspects, the business framework around them and the social interaction with the park’s surroundings form an extensive socio-organisational web [7].
A list of drivers and opportunities of EIPs is presented, as adapted from [13], [12], [7] and [14].
While the ideological basis for the development of an EIP can be environmental, the most important driver for a successful project is the financial gain. Business performance and environmental performance should therefore both play an important role in the design of the park and inter-firm cooperation will only then offer advantages compared to regular (more wasteful) business park design.
Two ways of thinking exist as to whether EIPs should be artificially engineered or self- organised. The first perspective focuses on the technical feasibility of the inter-firm linkages, while the second believes that organic growth of connections will result in Chapter 1. Introduction 4
enhanced ownership and a more resilient system. What is most important is that the system is fully integrated in its surrounding and that it possesses sufficient flexibility to survive in a dynamic economic environment.
Ideally, the initiative for inter-firm cooperation should come from the firms themselves and not from any other authority. The role of the park management should be to facilitate information flow between the firms and to provide a platform for open com- munication.
Physical proximity and complementary profiles in terms of energy and materials is important and direct economic competition between firms should be avoided [15]. In any case, a level of trust between the participants is crucial.
An active participation of the firms in collective projects ensures continuity and sus- tained ownership. Participation from additional stakeholders such as other companies, the local community, environmental organisations, as well as technical experts from different fields can be a strong asset in the success of a project.
During the planning phase of the site, the focus should be primarily on utility sharing. The subsequent economic and environmental advantages will convince the companies of the benefits of symbiotic exchanges [12]. In a second instance, energy, water and material waste exchanges can be considered. Finally more company specific and economically challenging projects can be developed.
The most famous example of a successful EIP is the Kalundborg Park in Denmark. It is the first realisation of industrial symbiosis that was published in literature [3]. Nine different firms, including a coal-fired power station, an oil refinery and a pharmaceutical plant are connected through a network of eleven physical linkages. Remarkable is that the inter-connections were not conceived from the design of the park but instead grew gradually through independent and economically driven actions [16]. The Asnaes Power Station has a very central role in the system, which shows the fundamental role of energy supply in EIP design. The plant provides electricity as well as heat and process steam to its industrial and residential neighbours. The power station is still coal fired but a transition to biomass is planned in the near future [17].
While there are inherent advantages for interactions between firms, technical, economic, in- formational, organisational and regulatory barriers can be identified [12].
The main limitation of EIP parks is the potential for system fragility. If one of the park’s main firms decides to re-locate, other companies might need to find other sources for their raw materials thus affecting the functioning of the entire chain. Electrical power however, as opposed to physical primary resources, is homogeneous independently of Chapter 1. Introduction 5
its source, and therefore collective power supply is less affected by this fragility. Dis- tribution systems though will be sized and reinforced according to the type of installed generation, which means that a change in the production might result in poor power quality and reliability. This is worse in the case of heat exchange, because quality and temperature are very important parameters for process heat. One way of mitigating this vulnerability is a diversification of sources [13].
Another limitation is that the price of inputs and output, such as electricity and other energy forms, vary over time. This means that the profitability of inter-firm projects is subject to uncertainty. Furthermore, low market prices for (non-renewable) energy sources means that companies often do not have enough incentives to invest in collective renewable energy projects.
In the energy sector the legal framework is often a hindrance. Administrative barriers, public opposition and ever-changing or non-existent legislation create an unfavourable environment for the development of inter-firm energy supply. This limitation is more broadly assessed in chapter 2 on microgrids in the Belgian market.
1.1.3 Inter-firm energy supply
The high investment cost and significant green house gas emissions in industrial energy supply mean that sharing facilities between firms is a promising approach [18]. The collective nature of electrical distribution means that several advantages of an inter-firm power grid can be identified [18].
Liberalisation of energy markets might open up new opportunities for inter-firm energy projects. Using an existing installation to participate in new market mechanisms such as spot markets and ancillary services can lead to financial gains.
Bundling power production and demand can flatten a site’s power curve, which means that the size of the grid connection can be downscaled.
Advancements in terms of online communication and power system control allow an increase in the implementation of demand response programs.
Economies of scale can mean that joint, inter-firm projects can be carried out more cost-efficiently. This way a diverse set of technologies can be installed at a lower cost, which will lead to a more integrated power system.
It is important to note the potential of combined heat and power technology. On-site linking of the electric and thermal grid allows an increased share of renewables without jeopardizing stability [19]. Chapter 1. Introduction 6
Notwithstanding these opportunities, optimising an on-site power system is not an unam- biguous task because of the multitude of involved stakeholders [9].
1.1.4 Industrial microgrids
An industrial parks power system is characterised by localised electrical and thermal loads, high energetic consumption and spatial opportunities for the integration of renewable energy sources. For these reasons, the prevailing paradigm for electric power distribution on an eco-industrial park is that of a microgrid. According to Lasseter [20], the microgrid concept assumes a cluster of loads and microsources operating as a single controllable system that provides both power and heat to the local area. The components in this energy cluster function independently in possible interaction with the surrounding macrogrid [21]. This results in enhanced local control, which can lower the cost of energy distribution, aid the integration of renewable sources and thereby reduce green house gas emissions [22]. Furthermore, the possibility of islanding improves the system security of supply by detaching the microgrid’s reliability from that of the macrogrid.
Broadly following the categorisation of the International Energy Agency [23], five kinds of flexibility resources can be used to supply and balance the microgrid’s variable energetic demand: dispatchable power plants, non-dispatchable power plants, demand side management and demand response programs, energy storage facilities and interconnection with adjacent grids. A set of the most interesting technologies for a hybrid power system in Flanders is selected.
In their white paper ’A Journey to Green Energy’ [24], Flemish distribution system operator (DSO) Eandis confirms that the most important renewable technologies in Belgium are and will continue to be solar and wind power. In 2015, 15% of electricity in Belgium was produced by renewable sources, with 4% of netto production by solar power and 6.5% by wind power [25]. The potential of PV and wind energy depends strongly on the solar radiation and the wind speed profile at the considered site and production is therefore inherently variable in time [26].
In a power system, renewable sources with intermittent behaviour can be combined with dis- patchable units in order to maintain the system balance [24]. Cogeneration, the simultaneous generation of heat and power, has a high flexibility and can therefore allow the integration of a large share of fluctuating electricity sources [19]. A combined heat and power (CHP) plant significantly increases the overall fuel efficiency compared to separate production and can
therefore lead to an important reduction in CO2 emissions [27]. At the moment the Flemish government has a certificate-based subsidy scheme for CHP installations that shows their Chapter 1. Introduction 7
interest in this technology. The share of electricity produced by cogeneration in Flanders is expected to increase [28] and therefore this technology is chosen as a component in the microgrid.
Historically, the inability to store electricity laid the foundation for the current electrical power system; the instantaneous balance between production and consumption is one of its building blocks. However, recent development in electrical storage technologies means that storing electricity has become more viable, even without large infrastructure such as a pumped hydroelectric energy storage [23]. Different aspects of microgrid flexibility (load levelling, peak shaving and seasonal storage, etc.), can therefore be offered using on-site electrical storage [29]. For these reasons, this study will assess the possibilities for electrical storage in industrial microgrid design.
Nowadays, through the introduction of the smart grid concept, the greater need for flexibility and the active participation of the end-user, demand can be seen as a new kind of resource [30]. Demand side management (DSM) is the set of methods that aims to assure grid reliability through active participation of the costumer in the electricity market [31]. Instead of changing the production side, it intents to modify the consumer’s load profile in response to financial incentives [31]. Demand management programs in a microgrid context can bring significant changes regarding economy, reliability and flexibility [32] and will therefore be included in this study.
The point of common coupling (PCC) is the connection point where the microgrid is con- nected to the macrogrid. While stand-alone microgrids exist [33] [34], dense electrification in Flanders means that a grid-connection can be installed at a relatively low cost. Connecting the microgrid to the utility grid makes bi-directional power exchanges possible, which allows electricity imports or exports in order to meet an excess or surplus of local production.
Figure 1.2 gives a schematic overview of the grid topology according to the technologies that were listed above. In chapter 3 each of the technologies is discussed at length and a microgrid optimisation model is developed. Chapter 1. Introduction 8
Figure 1.2: Microgrid topology Chapter 1. Introduction 9
1.2 Eiland Zwijnaarde
Eiland Zwijnaarde is an industrial park in the region of Ghent that is currently under develop- ment. It aims to accommodate a combination of high-tech companies, laboratories, university buildings and water-based logistics. It is located on an island bordered by the Ghent Canal ring in the north, the river the Scheldt in the East and the Tijarm in the south and west. An aerial view of the site is shown in Figure 1.3.
The site is being developed by Waterwegen en Zeekanaal NV and NV Eiland Zwijnaarde. NV Eiland Zwijnaarde is a collaboration between the Ghent city developer SOGent, provincie Oost Vlaanderen, and two private partners. Ghent university is expected to be one of present institutions on the park.
Figure 1.3: Map of the site Eiland Zwijnaarde
Historically the site was a dumping place for toxic waste from a nearby textile factory. At the end of last century it had become a so-called brownfield and was remediated and leveled in order to prepare it for new development. Chapter 1. Introduction 10
Waterwegen en Zeekanaal NV is the owner of the site’s northern part and is planning to establish a water-based transport and logistics hub. The rest of the lands belong to NV Eiland Zwijnaarde, a public-private partnership that was created to develop the site’s southern part into a technology campus. The project is a collaboration with Ghent University, under the joint venture Tech Lane Ghent.
1.2.1 Eiland Zwijnaarde as an eco-industrial park
The City of Ghent, driven by the green-socialist coalition in power since 2012, has high ambitions regarding sustainability. By 2050 it aims to be climate neutral, which means that the entire Ghent region should have no negative impact on climate change. This is to guarantee the city’s quality of life. A climate plan (Klimaatplan) was ratified in 2015 to outline the actions that must be taken to reach this goal.
For business parks in Ghent the climate plan states that all means will be employed in or- der to assure sustainability, from their design until their decommissioning [35]. Specifically, reduction in energy consumption, local renewable production and rational use of energy and materials are stimulated. Furthermore, collective energy projects such as district heating will be facilitated. These sustainable principles are upheld in the development of Eiland Zwijn- aarde and the sustainability ambitions were further developed in concrete vision documents.
1.2.2 Vision of project
Two concepts play a main role in the development of the project’s vision: centralisation and sustainability. Centralisation is to what extent the firms in the park are collectively organised. Sustainability depends on how far-reaching the commitment to sustainability will be.
Two visions were developed in detail as part of project’s design process: Centralised - Maximal and Decentralised - Moderate. These are shown and explained in Figure 1.4.
Energy
In terms of energy, the two scenarios translate into the following objectives.
Central-Maximal The Central-Maximal approach aims to maximally take advantages of possible synergies between the inhabitants of the parks. The site is climate neutral and completely self- sufficient in terms of energy on a regional level. Employees, as well as residents of surrounding residential areas, participate in projects on sustainable energy supply. The Chapter 1. Introduction 11
Local organisation Centralised Decentralised
Intens and ambitious cooperation between Maximal public and private parties.
Sustainability
An unbinding guidance of the private partners Moderate by the public party
Figure 1.4: Development scenarios for Eiland Zwijnaarde
central park management’s list of responsibilities is extensive and many decisions on sustainability issues are taken collectively on a park level. A code of conduct makes sure that the park’s residents comply with its sustainability principles. The ambitions regarding energy are an increase of locally produced power from 20% at the site’s startup to 100% by 2050. To this end, there will be large scale wind power and solar PV, complemented with electrical storage. Later in the project a cogeneration unit can be installed that is connected to a thermal grid. BREAAM certification (see section below) is the general guideline for the energy performance of buildings, with a focus on passive construction and maximal insulation. There is a centralised and integral energy management system that regularly carries out energy audits. The electric distribution grid is designed taking into account the renewable production and a high percentage of electric vehicles.
Decentral-Moderate For the Decentralised-Moderate scenario, the ambitions are along similar lines. However, there is less drive to carry out joint projects and the final goals are slightly less ambitious. Instead of 100% self-sufficiency by 2050, the Decentral-Moderate approach aims for 80%. Shared facilities such as storage and generation are considered but on an individual building-level instead of a park-level. A central heating network is absent, but inter-firm heat sharing between two or more firms can be considered according to the opportunities that arise. Chapter 1. Introduction 12
BREAAM certification
The Building Research Establishment Environmental Assessment Method (BREAAM) is an evaluation method used to assess the sustainability of buildings and regions. Building and park owners, as well as city councils, governments and project developers can choose to certify their building using BREAAM in order to show their dedication to sustainability.
The certification is not limited to energy savings and efficiency, instead it attempts to make a wholesome assessment according to the following categories:
Responsible management
Synergy with the surroundings
Sourcing of water, materials and energy
Spatial development
Well-being and prosperity
Regional climate and environment
Under each categories there are different topics for which credits can be earned according to sustainability criteria. These credits are awarded according to evidence-based proof.
Trias Ecologica
An important basis for the BREAAM certificate are the Trias Ecologica and its application on energy, the Trias Energetica. Both these strategies for sustainable development were developed by Kees Duijvestein at the TU Delft.
The Trias Ecologica is a way of acting and thinking in the field of sustainable design. The fundamental idea is to limit the input of energy, materials and water during the construction process of a building. Once it is built, the Trias aims to limit the building’s outflow of energy, materials and water. Concretely this means that renewable sources should be used rationally during construction, and that waste should be avoided or reused during its lifetime [11].
The application of the Trias Ecologica on the energetic design of a system results in the Trias Energetica, as shown in Figure 1.5. It consists of three strategic measures to sustainably de- velop and operate energy systems. The first step limits the system’s energetic consumption by eliminating losses and increasing the energetic efficiency. Secondly, consumed energy should be generated using sustainable and renewable sources. Thirdly, if non-renewable sources are inevitable, they should be used at the highest efficiency [36]. Chapter 1. Introduction 13
Reduce energy demand
Employ renewable sources Clean use of fossil fuels
Figure 1.5: Trias Energetica [36]
1.2.3 Current state of affairs
In order to learn about the recent advancements in the development of Eimand Zwijnaarde, an interview with Mieke Gevaerts was conducted. She is the site’s project manager at SOGent.
The fact that both public and private stakeholders are involved in the project makes it harder to pursue a collective vision. There is political motivation to develop a sustainable industrial park that is progressive in terms of energy, mobility and ecology. However this is opposed by the very market-driven industrial park development business. Furthermore, the development process consists of many stages and in every stage each stakeholder pursues its own interests.
At the moment the project developers are trying to translate the project’s vision documents into an attractive business environment for firms. Agreements with several companies are under way and there is a large commitment from Ghent University to move to the park. As identified earlier in this chapter, active participation of the firms is a determining factor in the success of inter-firm cooperation and it is therefore important that they are involved from an early stage. However, for Eiland Zwijnaarde, the commercialisation of the project has meant that some compromises have been made on the original vision documents.
The code of conduct that would bind the companies to follow certain principles regarding sustainability will not be implemented since it would over-complicate the sales contract. In- stead, obligatory BREAAM certification for all buildings will be used to assure sustainability. In a sales contract it is not possible to include future development regarding construction and technology. BREAAM certification, despite its high price, has the advantage that it is updated regularly and will therefore never be outdated.
Several studies have been done on the design of the site’s energetic infrastructure. Because of the uncertain demand profiles and a possible slow ingress of companies, the proposed collective infrastructure projects such as a heating network and on-site cogeneration have been found Chapter 1. Introduction 14 economically infeasible. Instead the traditional electrical and gas grid will be installed. The potential of heat exchange with a nearby incineration facility was assessed but the fact that it can take until 2030 before the park is fully filled means that the initial thermal demand will be too low for profitability. In terms of renewable generation there are still plans for large scale wind and solar PV. At the moment one wind turbine is being certified and building-based geothermal energy is considered as a potential renewable source.
A special-purpose vehicle SPV Energie was founded to facilitate future projects regarding energy. Its role is not completely defined and can be adapted according to the site’s devel- opment. It could for example finance collective energy projects or be used as a platform for inter-firm communication on energy supply. The structure also allows participation of other stakeholders such as surrounding inhabitants or energy cooperatives. There have been talks with Energent, a Ghent-based energy cooperative, but at the moment they are not involved in the development of the site.
The different development scenarios for Eiland Zwijnaarde can be classified according to the EIP concepts defined in the previous section. In the Maximal-Central scenario, far-reaching collective measures were proposed that would affect not only the technical but also the social aspects of the park. This would lead to a very integrated industrial eco-system with strong links between the firms. The result would most closely correspond to a sustainable industrial park. The Moderate-Decentral scenario still has a focus on inter-firm cooperation but allows for more individuality. It would therefore result in an eco-industrial park. In the development scenario that is actually being carried out, the firms can choose their own role in the park’s sustainability. While the firms have to fulfil certain sustainability criteria, the final degree of sustainability will strongly depend on their own initiatives. Their motivation will be decisive as to whether Eiland Zwijnaarde will result in a green or eco-industrial park. Chapter 2
Microgrids in the Belgian electricity sector
The Belgian energy market is a complex picture with many different actors. In order to properly assess the profitability of the microgrid’s components as outlined in the section above, a general overview of the Belgian electricity sector is necessary. The aim here is to identify the roles a microgrid can play within the Belgian power sector and to assess the opportunities new technologies present in an economic context.
2.1 The liberalised electricity market
The recent transition from a state-owned, vertically integrated energy sector to a liberalised market has changed energy planning and introduced various new market roles in the elec- tricity market. While before a single company could operate across the entire value chain of energy, different EU liberalisation directives have disaggregated the market. Concretely this has caused a separation of the electric power business and the operation of the network infrastructure (distinguishing competitive and non-competitive parts of the industry). It al- lows access to energy infrastructure to third parties and implies a liberated supply side of the market with unrestricted change of supplier for the consumers [37]. Independent regulators, in Belgium the VREG, CREG, etc., were created to monitor the sector.
The aim of the transition is to lower energy prices due to increased competition and develop customised energy services. However, in the energy sector there is a long time between the initiation and completion of projects, which can be hard to reconcile with the liberalised mar- ket’s favour for high returns and quick payback times. Furthermore, for inter-firm electricity production, the strict unbundling of production and distribution as well as the prohibition of one-to-one electric connections between firms pose restrictions on the possibilities for electrical energy clustering [38].
15 Chapter 2. Microgrids in the Belgian electricity sector 16 Since the begin- 8 Technical and commercial roles in an electricity market with full retail competition ning of the new Generation Transmission
Power stations millennium, many Transmission company Sales Distribution countries have System operator Distribution companies moved towards Balance responsible Network access strong support Wholesale Spot market/PX management Metering Energy for solar and wind service provider Customer energy. Financial market (ESP) Aggregator
Ownership Money Energy Product Service
Figure 2.1: Lay-out of a liberalised electricity market [39]
9 The huge growth of wind and solar energy of recent years has created new challenges.
Figure 2.1 shows the market roles in a liberalised market. The different actors in the electricity sector are listed below [40] [41] [42].
Production: Electricity generators are the first players in the value chain. The power they produce is either directly consumed on-site (auto-consumption), or injected in the transmission or distribution grid. The producers sell their electricity on one of the power exchange markets or through bilateral contracts. They are in direct communication with the balance responsible party and are able to provide ancillary services to the grid according to their capacity.
Transmission: The transmission system consists of all national high-voltage lines and transmits elec- tricity from the generators to either high-voltage (industrial) consumers or to the distri- bution grid. The transmission grid also allows electricity imports from abroad to satisfy national demand. In Belgium the transmission system operator (TSO) is Elia. Its role in the Belgian market is to operate the electrical system and infrastructure, which includes managing the balance between generation and consumption and maintaining the high-voltage facilities. Because of its fundamental role in the sector it acts as the market facilitator and facilitates access to the grid for producers and consumers. This energy can be provided at a price below – Distributed generation is increasing way it aims to promote free competition between the market players. It functions in a that paid by consumers on the low volt- mainly because of photovoltaics and natural monopoly and is therefore subject to monitoring by regulatory bodies. [43] age network. Because photovoltaics dis- combined heat and power and will plays an almost linear cost structure cause a significant share of genera- (without significant economies of scale in tion to be covered by a very large the investment costs) it is having a fun- number of small units. damental impact on the economics, and – Volatile production from wind and hence also the structure, of the electrici- solar energy is leading to faster and ty supply. The main characteristics of larger supply-side fluctuations of only this impact, from a technical systemic limited predictability. perspective are: – A greater geographic separation These three changes have technical im- between generation and consumption plications in all aspects of the supply and is introduced to systems previously use of electrical energy ➔ 10. Two chang- built mainly around fossil fuels or es are particularly noteworthy: the grow- nuclear energy and which previously ing importance of long-range and high- balanced consumption and genera- performance transmission networks and tion on a regional level. This develop- the integration of highly-distributed ele- ment is driven primarily by strong ments, both on the production side and location-dependent primary energy on the consumption side (smart man- sources such as wind and water. agement of consumption).
52 ABB review 4|14 Chapter 2. Microgrids in the Belgian electricity sector 17
Distribution: The distribution grid operator (DSO) operates the medium to low-voltage grid and transmits electricity from the sources to the end-users. It is responsible for the exploit- ation, maintenance and development of the distribution network and it functions just like the TSO in a natural monopoly. The DSO monitors the consumers’ consumption and passes on the information to the supplier. In Flanders different DSOs exist such as Eandis and Infrax.
Suppliers and power exchanges: Suppliers sell electricity to households and companies. They can either buy the electri- city from a power exchange or obtain power through bilateral contracts with producers. In a liberalised market suppliers can offer different types of contracts and products, de- pending on the costumers’ demands. Examples are ’100% renewable’ electricity, guaran- teed domestic production, etc. Suppliers are responsible for the billing of the consumed electricity. This consists of a volumetric tariff for consumed power, a fee for the network costs as well as the taxes and levies. They distribute these components to the respective market actors. They additionally charge a margin for their own operational costs. Large consumers and traders rely on the Belgian ICE power exchange to anonymously trade electricity. Different markets exist. On the Futures Market electricity can be traded 3 years ahead. About 10% of all traded electricity is sold this way. The day ahead market sells electricity for the next day with a varying price per hour. On the intra-day market electricity can be sold and purchased on a continuous basis until five minutes before delivery. Each of these markets has its own characteristics with widely varying prices that depend on the TSO zone as well as the time of the day. Market players can also decide not to trade through the power exchange and instead exchange electricity directly via bilateral contracts. This is called ’over the counter’ (OTC) or wholesale exchange. The bulk of the electricity is sold this way. OTC allows customised products to be sold and sales generally happen faster than through the power exchange. The balancing market is the market that facilitates grid balancing and is not part of the power exchange. It consists of several mechanisms to maintain the balance of the grid. These mechanisms are called ancillary services and use financial incentives to convince consumers or producers to adjust their production/consumption profile according to the grid’s needs.
Balance responsible party: The power produced at every point in time should equal the power consumed. This balance is maintained by Elia in several ways. Chapter 2. Microgrids in the Belgian electricity sector 18
At every access point in the transmission grid there are access responsible parties (ARP) appointed by Elia. They can be a producer, a consumer, an energy supplier or a trader and are responsible for the balance of power injection and extraction at that point in the transmission grid. They pay a fee for any imbalance between injections, import and purchases on the one hand, and offtakes, export and sales on the other. When the ARP is unable to deliver the balancing service, the TSO can fall back on ancillary services to maintain the balance. The offending ARP is charged for this.
Aggregator: Decentralised production units and loads can be grouped into a single entity when interacting with the electricity market [44]. This is called aggregation. An aggregator is a company that functions as the link between distributed energy sources and end-user on one hand and other actors such as the DSO and TSO on the other. Because of enhanced control they offer, aggregators can play an important role in the transition towards a high fraction of installed non-dispatchable generation.
End-users: The end-users consume electricity to fulfill their needs. Consumers can be large indus- trial sites, office buildings, households, etc. These are either connected to the distri- bution grid or directly to the transmission grid. Although earlier end-users were often considered as passive consumers, at the moment there is a strong shift towards active market participation. ’Prosuming’, end-users that consume as well as produce, and de- mand management programs that adapt consumption according to the market’s needs, allow end-users to take an active role in the market.
2.2 Price of electricity
The price of electricity is not just a compensation for the cost of generation. The total price a consumer pays for the consumption of electricity also consists of the network costs, taxes and governmental levies. Each of the actors outlined above will have their share in the final electricity bill. Figure 2.2 shows the breakdown in percentage of each of the different components. The values are based on a study done by the VREG for an industrial consumer with an annual consumption of 50 MWh. Chapter 2. Microgrids in the Belgian electricity sector 19
2% 7% Commodity price
Distribution cost 41% Transmission cost 50% Levies
Figure 2.2: Breakdown of the price of electricity [45]
The commodity price and the distribution cost account for more than 90% of the total cost. It is remarkable the distribution cost is higher than the energy cost. The transmission cost is much lower than the distribution cost. VAT is not included.
2.2.1 Commodity price
Electricity is a commodity in the sense that once it is produced its source becomes irrelevant. However, the fact that large-scale electrical storage is still unprofitable means that electri- city cannot be stored and that grid stability has a very strong impact on the instantaneous electricity price. This means that the market price of electricity can vary between one value and its tenfold within the same day, and this volatile behaviour is only aggravated by the increased integration of intermittent renewable energy sources.
In the section above the different markets in Belgium are briefly outlined. Figure 2.3 0.45 shows the Belpex day ahead market price for 0.4 the year 2015. The mean value is around 0.35 50 e per MWh of electricity, and there are 0.3 peaks of up to 450 e per MWh. This dy- 0.25 namic shows that the time of consumption 0.2 0.15 can have an strong effect on the final price 0.1 paid for electricity, and this is particularly 0.05 important for large consumers. 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 hour
Figure 2.3: Hourly Belpex price in 2015 Chapter 2. Microgrids in the Belgian electricity sector 20
2.2.2 Transmission tarrifs
The transmission tarrif is charged by Elia and consists of the following components [46].
A. Connection The connection cost is the cost of construction for the grid connection of a specific site. It is an important parameter in this study since an increase in the microgrid’s self-sufficiency could imply a lower connection cost. However, in Elia’s pricing scheme, most terms in the calculation of the connection cost only depend on the voltage of the connection and not explicitly on the peak power consumption[46]. The only term that does depend on the maximum power is the transformer capacity. In Elia’s calculations the transformer cost depends on its capacity according to Equation 2.1.
0.75 MVA C = C0 0.75 · (2.1) " MVA0 #
The cost of a 36kV-30kV/MT transformer with rated power 25 MVA is 34 810 e. This means that a transformer with rated power 10MVA would still cost 22 231 e. It is clear that a decrease of just 12 579 e over the lifetime of the microgrid is rather insignificant compared to the investment cost of a couple of million Euro for distributed generation units. While it is not impossible to find financial incentives for a smaller grid connection for microgrids, these will most likely not be on the transmission tariff level.
B. Management and development of grid infrastructure Cost for the monthly and yearly peak in offtake and the power put at disposal.
C. Operation of the electricity system Cost for the management of the electric system and the offtake of additional reactive power.
D. Compensation of imbalances Cost for power reserves, black-start and the maintenance and restoring of the residual balance of the individual access responsible parties.
E. Market integration
F. Public duties Public duties include the connection of the off-shore wind power parks in the north sea, the federal green certificates and promoting the rational use of energy, etc.
G. Taxes and levies Chapter 2. Microgrids in the Belgian electricity sector 21
In practice many of these tariffs depend on the voltage of the connection with different costs between 380/220/150 kV networks and 70/36/30 kV networks. The tariffs for networks at higher voltages is generally lower. If the yearly transmission cost for a grid connection at 30 kV is calculated for a large industrial consumer with a yearly consumption of 40 TWh and a yearly and monthly peak of 7MW, the transmission cost is 12.47 e per MWh. The main cost of the transmission is the development of the transmission infrastructure, which is 33% of the bill. It is important to note that the operational costs of the TSO, items B, C and E in the previous list, make up only half of the price paid for transmission; the cost for public duties and taxes and levies make up the other half. This shows that policy has an important impact on the total transmission bill.
2.2.3 Distribution tarrifs
The distribution tariff consists of 3 aspects : a cost for power put at disposal, tariffs for system management and a metering cost [47]. This breakdown is analogous to the transmission cost. However, it is much harder to estimate a realistic distribution bill because the cost for distribution is much more heterogeneous depending on the geographical location of the consumer.
Just like transmission costs, distribution costs must be based on actual costs and are mon- itored by regulatory bodies. DSOs operating in urban regions will have much lower costs per amount of transported energy than those that operate in rural areas. In general it is possible to estimate that the distribution will be about half of the electricity bill. This is significantly more than the price paid for the transmission network, and even much higher than the com- modity price paid for electricity. A decrease in the DSO’s operating costs could therefore lead to a significant decrease in the total electricity bill. While this is not the focus of this study, it would be an interesting topic of research to assess the possibilities of adapting a microgrid’s distribution grid to its loads and generators in order to find the minimal size that guarantees operation. Auto-consumption and demand shifting could be employed to minimise the use of the distribution grid and lead to a smaller investment cost.
2.2.4 Industrial energy contract
Table 2.1 gives an overview of the different components in an industrial energy contract in Flanders. It was generated using the VREG’s VTEST tool [48] based on prices of May 2017. The prices are an estimate for an annual consumption of 10 MWh in the region Ghent with a monthly and yearly peak of 7MW. It is important to note that these prices are for a small industrial consumer connected to the low voltage grid. The specific energy cost per kWh Chapter 2. Microgrids in the Belgian electricity sector 22 gives an indication of the relative costs.
The result confirms the findings of the sections above. The commodity price is around 60 e/MWh, which is a little above the 2015 mean Belpex price. The network costs are more than half of the total cost. The total price is around 240 e/MWh. Note that VAT is not included.
Table 2.1: Example of an industrial energy contract
Calculation (excl. VAT (21%)) ENERGY COST ce/kWh kWh e606.80 Annual fee 25.00 Energy component - Single day-time rate 3.78 10 000 378.00 Costs for green power 1.82 10 000 181.80 Costs for CHP 0.22 10 000 22.00 NETWORK COST e1503.58 Distribution 1331.33 Transmission 172.25 LEVIES e353.12 Contribution to energy 19.26 Federal contribution 34.11 Degressiveness Federal contribution -0.00 Contribution to the Energy Fund 299.75 TOTAL 24.64 10 000 e2463.50
2.3 Energy policy in Flanders
An overview of European climate and energy policy up to 2013 relevant to industrial energy clustering can be found in the second chapter of the ”Low carbon business park manual” writ- ten under the project ACE [11]. Energy policy in Flanders is based on the Decree on Energy (het Energiedecreet) which was drafted in 2009 and took effect in 2011. The implemented provisions are described in het Energiebesluit, published in 2010 [49]. Below a brief summary of Flemish renewable energy policy and the regional measures for industry are given [50].
The federal government’s competences regarding energy consist of the transmission grid and its tariffs, and the production of electricity excluding renewable sources and cogeneration. Their policy for renewable energy is therefore limited to an enhanced investment deduction, which allows companies to deduct up to 13.5% (in 2017) from their taxable profit for energy Chapter 2. Microgrids in the Belgian electricity sector 23 saving investments.
The Flemish government provides different compensation schemes for energy-saving measures.
Ecology premium The ecology premium aims to stimulate companies to make their production processes energy-efficient and environmentally friendly. It consists of a partial reimbursement of the made investment. Different sustainable technologies are ranked into four categories according to their ecological impact. Financial support for a project is granted according to the investment’s category.
Strategic ecological support For investments that involve state-of-the-art technology, strategic ecological support can result in up to 40% reimbursement of the investment. It is aimed for very specific applications that are not included in the Ecology premium.
Property tax deduction For renovated and newly constructed non-residential buildings, firms can deduct a per- centage from their property tax depending on the buildings E-level.
Green power certificates Electricity producers in Flanders can obtain green power certificates for electricity produced from a renewable source. Such a certificate of origin proves that a certain amount of power was produced using a renewable source and is issued by the VREG. A technology-specific banding factor determines the exact amount of energy that has to be produced to obtain one certificate. This factor depends on the evolution of the investment cost, fuel cost, etc. Certificates are awarded during the entire duration of the support period, which is generally 10 years. Green power certificates can either be sold to the DSO or to electricity suppliers. Suppliers must acquire a certain number of certificates according to an obligatory quota. Certificates, as opposed to investment support or feed-in tariffs, are less bureaucratic and allow free market mechanisms to come up with the most competitive solution.
Cogeneration certificates Cogeneration certificates are similar to green power certificates, but instead of being awarded for electricity production they are awarded per amount of primary fuel saved compared to a reference power plant and reference boiler. Just like green power certi- ficates they can be sold to the DSO or electricity suppliers.
Support for energy audits Different schemes exist to offer firms a free and individual energy audit to identify opportunities for energy efficiency improvements. Chapter 2. Microgrids in the Belgian electricity sector 24
Subsidies and support schemes for renewables aid and accelerate the transition towards a more sustainable power system. However the large decrease in investment cost for several renewable technologies that has taken place in the past years means that these technologies no longer need additional financial incentives to be economically interesting. Support schemes have already decreased and will continue to do so. Therefore they should not be a major consideration when planning a future power system and will not be further included in this study. Although from a developer’s point of view a decrease in subsidies could be considered as a loss of potential revenue, it is in fact a very positive trend. It means that the transition towards a sustainable power system will take place at no additional cost to society.
2.4CO 2 market
The EU emission trading system (EU ETS) was launched in 2005 to incentivise the reduction of CO2 emissions in different industrial sectors [51] . It operates according to a cap and trade principle in which companies from included industries have to compensate their emissions through emission allowances. The total annual emissions from these industries is limited and this cap reduces every year. It offers the industry flexibility in finding the most economical way of reducing their CO2 emissions. The ETS system specifically focuses on the power and heat sector because of its large share in European greenhouse gas emissions.
Allowances, which are the currency in this system, can be traded on the so-called CO2 market. They can be bought from other companies or granted by the government. While the system has been successful at effectively reducing CO2 across industry [52], there is the risk of carbon leakage: businesses move their production to countries with less stringent CO2 legislation.
Furthermore, the current low carbon price (around 5 e/tCO2) means that at the moment this systems offers little incentives to invest in carbon neutral technology. However, the price per tonne is expected to increase. In its study on trends in energy, transport and green house gas emissions (GHG), the European commission predicts that the ETS price will increase steeply in the coming years [53]. Their forecast is shown in Figure 2.4. This forecast will be
used in the microgrid model to economically assess the impact of a high ETS CO2 price. Chapter 2. Microgrids in the Belgian electricity sector 25 ) 2 100
80 ’13/tCO e 60
40
20 ETS Carbon price ( 0 2020 2030 2040 2050
Figure 2.4: Projection of ETS CO2 price
2.5 Grid structure and microgrid opportunities
The grid structure as it is now is predominantly tree shaped. Power is produced by large generating units and transported in one direction across the utility grid towards the consumer. This, however, is rapidly changing due to the global shift towards distributed generation and governmental incentives for renewable and cogeneration technologies [54]. Instead of a uni- directional power flow from generator to end-user, technologies such as solar power, wind power and CHP installations allow consumers to meet their own electrical demand. This results in the need for more integrated power systems that are able to accommodate the subsequent bi-directional power flows [55].
In the introduction, microgrids were identified as a technology with a large potential regard- ing the transition towards this new and sustainable power system. Its benefits in terms of efficiency, reliability, (cyber)security, power quality, and sustainability will positively affect the entire value chain of energy, from the producer to the end-user [56]. In a way they can be considered as fundamental building blocks of the smart grid [57]. However, there is no clear consensus on what its functionality is, how it should be controlled and what market entity should operate it. The sections below discuss these topics.
2.5.1 Functionality and economic incentives
The global trend towards regional power grids that extend beyond national borders can rein- force the advancement of microgrid technology because of its ability to integrate producers, consumers and ’prosumers’ [21]. Several types of microgrids with different functionalities can Chapter 2. Microgrids in the Belgian electricity sector 26 be identified. Campus microgrids focus on shifting away from energy intensive university facilities towards sustainably powered technology campuses [58] [59]. Because they are often located in densely electrified areas, these microgrids are generally connected to a macrogrid. Critical asset microgrids such as a hospitals [22] primarily make use of the enhanced reliab- ility to maintain electricity supply even during macrogrid outages. Community microgrids group households and allow large civilian participation in the integration of a high RES share. Island microgrids and remote microgrids [34] [60] [33] are the most natural type of microgrids since these power systems inherently function in islanding mode and are generally not connec- ted to a macrogrid. Microgrid technology can thus clearly be adapted to the sites’ demands regarding power quality and location specifics.
This study focuses on industrial and campus microgrids. Local control means that object- ives such as optimal market participation, flattening of the load and production curves and improved power quality can be pursued [57]. To the macrogrid, an industrial microgrid is a single controlled entity, which has the advantage that is able to provide ancillary services such as black start or voltage control. The flexibility to operate on- or off-grid has a positive impact on reliability; in case of an outage on macro level, the microgrid can maintain the electrical supply by going in ’islanding’ mode. This avoids the extra cost of an individual uninterruptible power supply (UPS).
Nevertheless, a barrier for microgrids in densely electrified areas is that the additional infra- structure can be a major investment compared to a conventional grid connection. As shown in subsection 2.2.2, the low additional cost of increasing the transmission grid connection means that it is hard to economically justify a power system that allows islanding. Although the way electricity is billed at the moment makes it hard to find profitable business models for microgrid technologies, new market mechanisms such as ancillary services can offer financial support for microgrids that actively participate in the power market.
In literature the financial feasibility of microgrid projects is assessed in different cases. Us- ing the DER CAM modelling framework, Hanna et al. [22] suggest that microgrids can be cost-effective compared to complete reliance upon macrogrid utility service. A comparison is made between a large commercial building, critical infrastructure and a campus. Local production offers advantages in the case of large consumers when on-site production covers both the electrical and thermal load. According to the authors, some fraction of renewable generation will always be part of the optimal generation mix. However, due to the low gas price, cogeneration based on natural gas generators is identified as the most robust technology in the microgrid design. The authors therefore warn that without correct policies this could create difficulties in reducing green house gas emissions. Sreedharan et al. [58] specifically look at the campus of University of California - San Diego to develop a case study on the operation of its microgrid. Three roles of the microgrid in relation to the overarching utility Chapter 2. Microgrids in the Belgian electricity sector 27 grid are specifically studied: participation in the ancillary service market, peak load shifting, and on-site storage of solar power. All three are found to be technically feasible and eco- nomically profitable given the right tariff structures. Just like Hanna et al. [22], the authors emphasise the importance of integrated optimisation of both the power and heat systems. Furthermore unit commitment optimisation on monthly and daily basis is key for economic and technical efficiency. Ancillary services can be taken into account in the optimal dispatch of the units, although this can imply an increased operating cost [61]. The economic feasibility of a microgrid’s participation in the ancillary market thus strongly depends on the utility’s remuneration. Furthermore the provided services (hourly ramping, load following, frequency regulation, etc.) are determined by the microgrid’s flexibility constraints.
2.5.2 Control
Microgrid control can be divided into two hierarchical control levels: a local primary control and a smart microgrd control [57].
The local primary control is an offline control system that aids the instantaneous grid stability and power quality. It functions much quicker than the secondary control. When the microgrid is in islanded mode, it is responsible for the reliability of the system. In grid connected mode it can improve power quality and voltage levels.
The high-level secondary control structure uses communication and metering to achieve tech- nical as well as economic objectives. This is the smart control layer. It allows optimisation of the microgrid in terms of demand management, unit commitment, fuel savings and power exchange with the utility grid. It can also facilitate incentive-driven primary and secondary ancillary services to the macrogrid.
The microgrid model developed in this study only takes into account the smart layer which maintains the hourly power balance in the system. Voltage and frequency balances are not considered. Different control strategies for the grids components were investigated and are presented in the chapter on Hybrid Power Systems. No further investigation was done re- garding instantaneous grid stability.
2.5.3 Ownership
Sanz et al. [62], Hanna et al. [22] and Asmus and Lawrence [63], amongst others, identify that new business models are necessary to allow the development of microgrids. The unanswered question is: Who develops, owns, and operates them. An issue with theoretically optimised Chapter 2. Microgrids in the Belgian electricity sector 28 power systems is that because of the unbundled value chain of electricity the visions of many stakeholders have to be aligned and actual implementation of the optimal microgrid design is a complex task [64].
The following paragraphs presents a non-exhaustive lists of four organisational strategies in terms of microgrid management on industrial parks that could exist within the Belgian energy sector. For each the advantages and disadvantages are discussed.
DSO DSO-developed microgrids can become a strong case since they have plenty of experience with developing and operating power systems [63]. Because of the changes in the energy sector, DSOs need to diversify their business model and convert the power grid into an active network [38]. Developing microgrids could become a part of their new role and a step towards a smarter power system. However, one main problem with DSO operated microgrids is that legally, distribution companies are not allowed to produce or store electricity because of their natural monopoly. While they can interfere in the microgrid control through market mechanisms, this is only an indirect form of authority. This means that an additional operator is necessary. Nevertheless this barrier, because of their historic role in the electricity supply chain they will definitely become an important actor in microgrid development.
Dedicated entity A dedicated entity within the park, such as a park-specific energy supplier, is another possibility for the implementation and operation of a microgrid. Many EIPs already have a central park management to maintain facilities and to offer an inter-firm com- munication platform. Energy supply could either become one of this organisation’s responsibilities, or a new dedicated entity could be created. In the case of Eiland Zwijn- aarde this was done by creating SPV Energie. Setting up a cooperative energy supplier is another option. This would allow firms in the park as well as surrounding households to invest together in their own, shared electricity supply. As identified in the chapter on EIPs, this collective way of project finance has advantages in terms of responsibility and can have a positive impact on the project’s success. A drawback to this method is that diverse stakeholders are involved and that they can be hard to align towards a common goal.
External entity An external entity, such as an energy service company (ESCO), could be interested in implementing and operating an industrial park’s microgrid. ESCOs are companies whose primary business is to develop, install and finance projects regarding energy optimisation of facilities owned by their customer [65]. They often work according to Chapter 2. Microgrids in the Belgian electricity sector 29
performance-based contracts in which the revenue they receive is based on the actual energy savings due to the new technology. Since an ESCO is not bound to a single industrial park, one company could operate several microgrids and specialise in this service. Furthermore, ESCOs offer the advantage that the park’s energy management is coordinated by just one stakeholder, which offers organisational efficiency.
Individual actors Another solution could be to allow each firm to invest in their own production. The produced electricity can then be mutually traded between the other firms in the park. Recent developments regarding blockchain technology [66] means that one to one fin- ancial interactions could become quick and reliable. This way a highly interactive local power market can be constructed [67]. A remarkable case study concerns a microgrid that is currently being constructed in Brooklyn. It will use a blockchain-based ledger that will allow renewable electricity to be traded between neighbours without interme- diaries [68].
A stakeholder analysis that depends on the microgrid’s overarching organisational structure can be part of the planning process [69] [70]. This way a project can be optimised from the perspective of the different actors. For each technology it can be decided whether a individual or collective approach offers the biggest advantage. Chapter 3
Hybrid Power systems
To get familiar with the different concepts touched upon in this study on renewable generation on industrial parks, a thorough literature study was carried out on modelling of hybrid power systems. The set of technologies that was deemed promising for a microgrid in Flanders, as outlined in the introduction, are analysed. Different ways of modelling are compared and finally one is selected to be included in the microgrid simulation as used in chapter 4. For several technologies a preliminary analysis is done to see what economic conditions must be met in order to reach profitability.
The technologies that will be analysed are:
Wind power
Solar power
Combined heat and power
Electrical storage
Demand side management
Determining the optimal size of renewable generation sources is in the first place an optimal sizing problem, and therefore the technologies above are studied from this perspective. Table 3.1 gives an overview of the 18 articles on optimal sizing of hybrid power systems that were studied. For each of these the table shows the technologies that were included in the microgrid. In chapter 4 the same list of articles is used to talk about the different evaluation functions that can be used for the optimisation.
The developed optimisation model finds the optimal global system configuration for an in- dustrial microgrid. However, it is important to keep in mind that in the development of an actual EIP the system will not necessary be centrally planned and each firm is responsible for their own production and demand. Each of the stakeholders will have their own interests in the optimal system design. The developed model in this study does not take this aspect into account; instead it focuses on the technical combination of technologies that offers the best performance.
30 Chapter 3. Hybrid Power systems 31
Table 3.1: Overview of technologies in studied literature
Solar Wind Battery CHP Grid DR Backup source power power storage Koutroulis et al. [71] ••• Li et al. [72] Diesel ••• Tafreshi et al. [73] Biomass •••• Fossati et al. [74] Micro turbine, •••• fuel cell & diesel Ismail et al. [75] Micro turbine ••• Zhou et al. [32] •••• Yang et al. [76] ••• Zhao et al. [33] Diesel ••• Abbes et al. [77] ••• Dufo-L´opez and Diesel & Bernal-Agust´ın[78] ••• hydrogen Katsigiannis et al. [79] Diesel, fuel cell, ••• biodiesel & hydrogen Tito et al. [80] ••• Gonz´alezet al. [81] ••• Bayod-R´ujula et al. [82] •••• Theo et al. [83] Biomass •••• Kamjoo et al. [84] •••• Li et al. [85] ••• Yang et al. [86] •••
3.1 Wind power
Making an accurate model to predict wind power production at a specific site depends on wind speed data and boundary-layer meteorology and is therefore a complex task. In the section below an overview is given of the modelling techniques found in literature and the way wind power will be modelled in this study. An economic analysis is done that only includes wind power in order to learn more about the process of optimal sizing. Chapter 3. Hybrid Power systems 32
3.1.1 Wind turbine power curve
Pinst.wind Power output
0 0 Cut-in Rated Cut-out Wind speed
Figure 3.1: Wind turbine power curve
In 16 out of 18 studied articles, wind power is included as part of the hybrid energy system (see Table 3.1). All use wind speed data at the studied site to estimate the momentary wind power production.
The theoretical power in wind with speed v across an area A is calculated using equation 3.1 [84]. 1 P = ρ · A · v3 (3.1) theory 2 The wind turbine power coefficient Cp links this theoretical value with the actually produced power and depends on the specific wind turbine characteristics. Figure 3.1 shows a typical wind turbine power curve. The turbine starts producing electricity when the wind reaches the cut-in speed. As the wind speed increases, the production increases according to the curve, up to the point of rated power. For wind speeds greater than the rated wind speed, power output remains constant until the cut-out point. The wind turbine will not produce power for wind speeds above this value for safety reasons. Modelling of this curve can be done in different ways. Koutroulis et al. [71] uses pairs of wind speed and wind power production in a lookup table to determine the produced power at each time step. Other authors such as Li et al. [72], Tafreshi et al. [73] use a cubic stepwise-continuous approximation formula such as Chapter 3. Hybrid Power systems 33 the one below to determine the wind power production.
i 0, v < vci 3 avi bP , v < vi < v i inst.wind ci r Pwind = − (3.2) P , v < vi < v inst.wind r co i 0, vco < v i With v the instantaneous wind speed, vci the cut-in wind speed, vr the rated wind speed and vco the cut-out wind speed.
Instead of the cubic correlation, some authors such as Yang et al. [76] use a linear estimation i of the power curve for wind speeds between v and v : P = P v −vci . Fossati et al. [74] ci r wind inst vr−vci i2 i uses a quadratic extrapolation method for this region: Pwind = α · v + β · v + γ.
In table 3.2 the details of the power curve are given for some examples of commercial wind turbine technologies that have been installed in on-shore wind parks in Flanders.
Table 3.2: Commercial wind turbines
Gamesa G114 Enercon E-82 GE Energy 2.5xl Rated power 2500 kW 2300 kW 2500 kW Cut-in wind speed 2,0 m/s 3,0 m/s 3,5 m/s Rated wind speed 10,0 m/s 12,0 m/s 13,5 m/s Cut-out wind speed 25,0 m/s 34,0 m/s 25 m/s
3.1.2 Wind speed data
The temporal and spatial variability and predictability of wind speeds make wind power one of the hardest renewable energy sources to predict [87]. Furthermore, hourly location- specific wind speed data is not easily available. One project that deserves mentioning is the European Meteorological derived high resolution renewable energy source generation time series (EMHIRES) project which uses the NASA atmospheric reanalysis dataset to estimate hourly wind power production over the past 30 years starting from historic wind speed data. The used NASA dataset has a time step of one hour and can therefore be used for hourly analysis [88].
According the World Meteorological Organization’s regulations, wind speed should measured at a height of 10m. To get the wind speed at hub height in order to calculate the power pro- duction from the turbine power curve, two different extrapolation laws can be distinguished: Chapter 3. Hybrid Power systems 34
Logarithmic law: According to the logarithmic law, wind speed varies in the vertical direction according to the formula u∗ z d v = ln − + ψ(z, z0,L) (3.3) h κ z 0 with u∗ the friction velocity, κ the Von Karman constant, d is the zero plane displace-
ment, z0 is the surface roughness and ψ is a stability term. This approach is used by Kamjoo et al. [84] to calculated the wind speed at hub height. However, because it assumes neutral stability conditions it is not able to reproduce the wind speed’s expected daytime variations [89].
Exponential law: The exponential law is another extrapolation method. h a v = v (3.4) h ref h ref with a an exponent that ranges from 1/7 to 1/4. While it is a simpler method, a is not constant and depends on the height and the surface roughness [90]. Koutroulis et al. [71] and Gonz´alezet al. [81] use this law to calculate wind speeds at hub hight.
60
Log law 50 Power law
40
30 H eight [m ]
20
10
0 6 7 8 9 10 11 12 13 Wind speed [ms -1]
Figure 3.2: Wind speed profiles based on the log and power law
Even though much has been written on how to model wind power from wind speeds, wind speed data for the region of Flanders is hard to come by or simply does not exist. For accurate Chapter 3. Hybrid Power systems 35 data for a particular location, on-site measurements are crucial [90]. Global wind speed data sets do exist, such as MERRA which was used in the above-mentioned EMHIRES project. However, extracting data in the form of time-series from the vast map-based reanalysis data- sets is cumbersome and time consuming. Instead, the Belgian TSO Elia publishes hourly time series of wind power production and predictions since 2012 which can be downloaded from their website. Figure 3.3 shows an example for the first week of January 2015. The aggregated dataset for on-shore wind power production in 2015 will be used for the simulations in this study. The hourly production is divided by the monitored capacity to obtain a time series of instantaneous capacity factors.
Figure 3.3: Historical wind production data
3.1.3 Dynamic economic analysis
A simulation is run to analyse the profitability of wind power in more detail. The aim is to optimise the installation size, taking into account a certain load profile and power exchange with the grid. It is essential to include the time component in simulations with non-dispatchable units such as wind power. This is because even though the renewable production over one year might equal the annual electric demand, big variations are possible within a smaller time step. In further simulations this method of simulation also allows to include the important dynamics of the electricity market.
The load profile is that of an industrial park in the region of Ghent with an average hourly consumption of 1 MWh (8760 MWh yearly). As shown in Figure 3.4, there is no strong seasonal dependency. The installed power of the wind turbine is varied between 0 MW and 5 MW. On-shore wind power in Belgium has a capacity factor of about 0.24 which means that for Pinst.wind = 4.1 MW the yearly produced wind power matches the year electric demand Chapter 3. Hybrid Power systems 36 exactly. The hourly production, as shown in Figure 3.5, is calculated by multiplying the installed power with the hourly historical capacity factors. The power exchange from the grid i i i for every time step can then be calculated according to the formula Pgrid = Pwind LPE. i − According to this convention, power drawn from the grid Pgrid is negative and power injected in grid is positive. There is no limit on imports and exports to the grid. An example of the power curves for an arbitrary week are shown in Figure 3.7. The price for electricity exchanged with the grid is varied between 50 e/MWh and 150 e/MWh, which reflects the price range for an industrial consumer connected to the medium voltage grid. The buying price is 105% of this value, the selling price 95%. These value are summarised in Table 3.3
Table 3.3: Simulation parameters for wind power
Pinst.wind: 0-5 MW Consumption: 8760 MWh/year
CAPEXwind: 1900 e/kWinstalled
OPEXwind: 0.02 e/kWh Grid price: 50-150 e/MWh Life time: 25 year Interest rate: 3 %
.
For each value of installed power and grid price, the cost of electricity (COE) is calculated according to the formula below.
LT OPEX Total cost = CAPEX + (1 + r)j j=1 X (3.5) Total consumption = LT · Annual consumption Total cost COE = Total consumption With CAPEX the intial investment cost, OPEX the annual operating and maintenance costs, r the interest rate and LT the life time (see Table 3.3).
Figure 3.8 presents the simulation results. At a high electricity price, increasing the amount of installed wind power from 0 to 5MW will cause the COE to drop from 109.7 e to 42.2 e per MWh of consumed power. For a low grid price this trend changes; the case with no installed wind power gives a COE of 36.6 e/MWh while 5MW wind power gives 54.3 e/MWh.
These results give a good indication of the behaviour of the wind power cost function and show that even though wind power has a very low operating cost per kWh, the high initial Chapter 3. Hybrid Power systems 37
1800 3500
3000 1600 2500
2000 1400 1500
1200 1000
500
Load profile (kW) 1000 0
Power exchange with grid (kW) -500 800 -1000
600 -1500 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 hour hour
Figure 3.4: Hourly load profile Figure 3.6: Hourly power exchange with the grid
4000 2000 P wind 3500 P 1500 grid LP E 3000 1000
2500 500 2000 0
1500 Power in kW
-500 1000 Wind power production (kW)
500 -1000
0 -1500 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 3000 3020 3040 3060 3080 3100 3120 3140 3160 hour hour
Figure 3.5: Hourly wind power production Figure 3.7: Power flows over one week
investment cost and the necessary imported electricity for balancing make a good sizing of the installation essential. One interesting case is when the electricity price is 70 e/MWh. The simulation results for this case are presented in Figure 3.9. The COE initially decreases for increasing wind power, to reach a minimum at Pwind between 1.5MW and 2MW. If the amount of installed wind power is increased further, the COE starts to rise. This shows that an optimal value of installed wind power exists; it is exactly this point that this study aims to find.
3.2 Solar power
Solar power has distinct characteristics compared to wind power. While it also depends on meteorological conditions, the spatial variation on a macro level is not as strong and solar Chapter 3. Hybrid Power systems 38
0.12
0.1
0.08
0.06
0.04
0.02
0.15 0.13 0.11 0.09 4000 5000 0.07 3000 0.05 2000 0 1000 P [kW] wind
Figure 3.8: 3D Simulation results for wind power production has a much more outspoken diurnal rhythm. However, clouds can cause variations on a very local scale especially during sunny days, which causes very sharp production peaks [89].
Below different modelling techniques are outlined. An economic analysis is carried out similar to the one for wind power.
3.2.1 PV-curve & Maximum power point (MPP)
17 out of the 18 articles that are studied include solar power in their hybrid power system. In general, three main ways of modeling are identified.
The first method used the photovoltaic (PV) current-voltage and power-voltage character- istics [71] [80]. It calculates the maximum power of the PV module depending on the solar irradiation conditions and ambient temperature according to the following formulas [71].
i · · Psolar = VOC ISC FF
VOC = VOC,ST C KV · TC − ◦ G (3.6) ISC = (ISC,ST C + KI · [TC 25 ]) · − 1000 ◦ Tnom 20 TC = T + − · G amb 800 The meteorological conditions herein are the global irradiance G and the ambient temperature
Tamb. The other variables are specifications of the photovoltaic cell: ISC is the PV short- Chapter 3. Hybrid Power systems 39
50.68 50 /MWh e
45 COE in
40 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Pinst.wind in MW
Figure 3.9: Simulation results for grid price of 70 e/MWh
circuit current, ISC,ST C is the circuit current under standard test conditions and KI the short-circuit current temperature coefficient. VOC is the open-circuit voltage, VOC,ST C the open-circuit voltage under standard test conditions, KV the open-circuit voltage temperature coefficient. Tnom is the nominal cell operating temperature and FF the fill factor. The variables VOC and ISC are thus the temperature- and irradation-corrected output voltage and current. The derived value for the power output Psolar has to be multiplied by the number of modules connected in series and parallel to find the installation’s total power output.
A second method involves the calculation of the PV’s power output depending on the irradi- ation G, the area A and a temperature corrected efficiency:
Psolar = ηpv · A · G (3.7) ηpv = ηr · ηpc · (1 β · (Tcell Tnom)) − −
ηr is the reference module efficiency, ηpc the power conditioning efficiency and β the generator efficiency temperature coefficient. Tcell is the cell temperature and Tnom is the nominal cell temperature [77]. This method and variations thereof are used in 6 articles.
In the two methods above it is important to note that the theoretically calculated output power is the power produced under a set of meteorological conditions given that the PV module is operated in its maximum power point. Equation 3.2.1 shows the power-voltage characteristic of a PV cell under constant radiation. Depending on the external voltage that is applied to the cell’s terminals, a different amount power is extracted. At a certain voltage
VMPP, this value reaches a maximum. Techniques for searching this maximum are called maximum power point tracking (MPPT). Chapter 3. Hybrid Power systems 40
MPP PMPP Power output
0 0 VMPP Voltage
Figure 3.10: PV power-voltage characteristic
A last way of modelling solar power production is by taking historical data directly as an indication for future production. This is done by Zhou et al. [32] and Zhao et al. [33] and will be used here. Figure 3.11 gives the solar power production in the Belgian province East-Flanders for the first week of January 2015.
Figure 3.11: Historical solar production data
3.2.2 Dynamic economic analysis
For solar power a similar analysis is run as the one for wind. The parameters are listed in Table 3.4. Except for the capital investment cost and the operational cost they are completely Chapter 3. Hybrid Power systems 41 identical. The installed power is varied between 0 and 5MW, and the grid price is varied between 50 e and 150 e per MWh. The interaction with the grid is simulated in the same way and the COE evaluation function as well. It is important to note that since the capacity
factor of solar power in Belgium is only 12%, a total of Pinst.solar = 8.42 MW is necessary in order to produce all of the annually consumed power.
Table 3.4: Overview simulation parameters for solar power
Pinst.solar: 0-5 MW Consumption: 8760 MWh/year
CAPEXsolar: 1570 e/kWinstalled
OPEXsolar: 25 e/MWh Grid price: 50-150 e/MWh Life time: 25 year Interest rate: 3 %
A comparison between Figure 3.13 and Figure 3.7 shows that the production profile of solar power is more season-dependent and therefore more predictable. A closer look at weekly power flows in Figure 3.13 shows that during the night the solar production is zero. Electricity has to be imported from the grid to meet the electric demand. During the day, the production increases to reach a peak around noon, after which production starts to drop. While at night electricity is imported, during the day in general there is a surplus and electricity is sold to the grid.
4500 4000 P solar 4000 3500 P grid 3000 LP 3500 E
2500 3000 2000 2500 1500 2000 1000 Power in kW 1500 500
Solar power production (kW) 1000 0
500 -500
0 -1000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 3000 3020 3040 3060 3080 3100 3120 3140 3160 hour hour
Figure 3.12: Hourly solar power production Figure 3.13: Power flows over one week
Figure 3.14 shows the 3-dimensional simulation results. As with wind power, for a high grid price an increase in the amount of installed power will cause a decrease in COE. However, in the case of solar power this decrease is not as pronounced; from 109.70 e to 93.11 e per MWh, Chapter 3. Hybrid Power systems 42 compared to 42.2 e/MWh for wind. Reasons for this are the lower power factor and the higher operational cost. For low grid prices, the COE will increase with increasing installed solar power from 36.57 e/MWh to 61.83 e/MWh.
80 79.89 0.12
0.1 /MWh e 0.08 0.06 75 0.04 COE in
0.02
0.15 0.13 5000 0.11 4000 0.09 3000 70 2000 0.07 1000 0.05 0 P [kW] solar 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Pinst.solar in MW Figure 3.14: 3D Simulation results solar power Figure 3.15: Simulation for grid price of 110 e/MWh
An interesting case is when the grid price is 110 e/MWh. These results are shows in Fig- ure 3.15. The optimal point is reached for an amount of installed solar power between 2 and 2.5MW with a COE of 79.89 e/MWh. This example shows that solar power needs higher grid prices in order to be cost efficient compared to wind power.
3.3 Cogeneration
The potential of cogeneration, or combined heat and power lies in that it is able to produce electricity and heat with higher primary energy resource efficiencies in comparison to separate heat and power generation [91]. An efficiency exceeding 90% is possible [92]. This is because the waste heat that is generated in the electricity production process is captured and reused to cover a heat demand. Figure 3.16 shows the general lay-out of a CHP installation.
There are many advantages to using cogeneration in a power system. Heavy investments in intermittent renewable power generation in recent years have created new requirements for the dynamic system reserve management of electrical grids. The short term balancing needs of power systems create opportunities for plants that can offer fast system reserves. Ancillary services markets will typically reward dynamic capabilities, such as fast starts, stops and load ramps. CHP plants can profitably participate in such markets provided that the power Chapter 3. Hybrid Power systems 43
Figure 3.16: Lay-out of a CHP system [93]
production can be operated independently from the heat production while maintaining high total efficiency at all times [94].
While CHPs are highly efficient, operate over a large load range and can provide the ne- cessary flexibility to assist the integration of non-dispatchable sources [94], there still exist some important barriers for its complete rollout. Colmenar-Santos et al. [95] identify 3 main reasons for this trend. Primarily, as long as energy efficiency is not valued more highly in the regulatory framework, the required additional investment for CHP installations will not be attractive for utilities. Furthermore, the unharmonised legislation in Europe makes it hard for players in one national market to enter another. Secondly, the current fuel and electricity price volatility poses significant risks for CHP project developers and cause higher prices. The financial attractiveness of cogeneration is inversely proportional to the spark spread1 and as long as the spark spread keep shrinking, cogeneration projects will have a hard time to take off. Finally, the long-term nature of district heating projects is hard to reconcile with the liberalised energy market’s preference for projects with short payback periods.
3.3.1 Cogeneration technologies
Table 3.5 presents an overview of the different types of CHP installations and their perform- ance characteristics. It can be seen that the electrical efficiency varies across a wide range, mainly depending on the size of the installation. Large systems are generally more efficient and the highest electrical efficiencies are achieved by fuel cells. While conventionally a high
1The theoretical gross margin of a gas-fired power plant from selling a unit of electricity, having bought the fuel required to produce this unit of electricity. spark spread = electricity price − (gas price × heat rate) Chapter 3. Hybrid Power systems 44
electric efficiency is preferred, a lower electrical efficiency will result in more heat that can be recovered to meet the thermal demand. Overall efficiencies are more or less equal across tech- nologies, between 70 and 80 %. The overall efficiency depends on the quality of the delivered heat.
Table 3.5: Overview of CHP technologies [96]
Recip. Engine Steam Turbine Gas Turbine Microturbine Fuel Cell
ηel 27-41% 5-40% 24-36% 22-28% 30-63%
ηtot 77-80% 80% 66-71% 63-70% 55-80%
Typical MWe 0.005-10 0.05-300 0.5-300 0.03-0.25 0.005-2 P-Q ratio 0.5-1.2 0.07-0.1 0.6-1.1 0.5-0.7 1-2 Part-load ok ok poor ok good Start-up time 10 sec 1 hr - 1 day 10 min - 1 hr 1 min sec 3 hrs - 2 days Fuels Natural gas All Natural gas, Natural gas, Hydrogen, Biogas, LPG, Synthetic gas, Sour gas, Natural gas, Industrial waste gas, Landfill gas, Liquid fuels Propane, Sour gas Fuel oils Methanol
Reciprocating engines Reciprocating engines have a high electrical efficiency and can be used in mobile mod- ules. It is a mature and financially attractive technology that maintains high efficiency under part load. Its relatively high ramp rate and rapid start-up capability allows it to be used as a UPS during grid failures. However, GHG emissions tend to be high and maintenance costs as well.
Steam turbine systems Steam turbines are often bigger than reciprocating engines and can run on a wide variety of fuels. Its design is adaptable to a big range of applications. It has OK performance under part load, however its ramping time is considerably higher than for reciprocating engines. The thermal energy is of higher quality, but the power to heat ratio is low.
Gas turbines The strength of gas turbines is that they can be made in a wide range of sizes, from microturbines to large central power stations, and that the released heat is at a very high temperature. Another advantage is that GHG emissions are low. However, the electrical efficiency is slightly less than for reciprocating engines and steam turbines. Gas turbines are reliable and have a long life time. Chapter 3. Hybrid Power systems 45
Microturbines Microturbines are small-scale gas turbines. They are discussed separately because of their distinct operational characteristics and applications. Because micro-installation are built on-site, thermal and electrical losses can be significantly reduced. They are mainly used for small-scale commercial and industrial applications.
Fuel cells The previous four systems all use direct combustion to convert the fuels energy first to heat and afterwards to electricity. Fuel cells do this differently; they use an electro- chemical process to convert the chemical energy of hydrogen directly into electricity. Many types exist. For example, a proton exchange membrane fuel cell transforms the chemical energy liberated during the electrochemical reaction of hydrogen and oxygen to electrical energy. The recovered heat is often at a low temperature. Fuel cells can use hydrogen from different sources and the operation is characterised by the type of electrochemical process that is used. Fuel cells have a very low noise generation, low GHG emissions and perform well under part load. However, capital investment costs are high because of custom production processes.
Because of their quick start up time, reciprocating engines are chosen to be included in the microgrid model. Instead of installing a separate emergency power system, a reciprocating engine cogeneration unit combined with electrical storage can have sufficient capacity and flexibility to replace a separate UPS.
3.3.2 Modelling and sizing
Modelling of a CHP is not unambiguous because of its two-dimensional interdependent set point, one for the thermal load and one for the electric load. Many variations in system design and control strategies make the optimal sizing of a cogeneration unit a complex task [97] and optimal sizing of a cogeneration installation in the context of microgrid design is therefore often studied separately. However, this way the interaction between dispatchable and non-dispatchable sources and the grid connection cannot be taken into account.
Only two of the mentioned authors in Table 3.1 include cogeneration in the hybrid power system’s design. Ismail et al. [75] note that very few authors on optimal sizing of hybrid power system take into account the cogeneration feature of micro-sources. The author presents a case study for a power system that includes cogeneration for remote communities. However, there is little focus on the dispatch of the CHP sources and no interaction with the utility grid. In Fossati et al. [74] the microgrid design consists of a wide range of technologies, including different dispatchable micro-source units such as a microturbine, a fuel cell and Chapter 3. Hybrid Power systems 46 diesel generators. Heat recovered from the micro turbine and the fuel cell is used to meet the grid’s thermal demand. However, the optimisation algorithm only treats the sizing of the battery storage system. For the cogeneration units only the unit commitment is discussed.
Below an overview is given of two particular types of sizing and dispatch studies that only focus on CHP installations. The first group of articles treat the general sizing aspects and therefore do not go into much detail on the instantaneous operation. These often employ a statistical or cumulative approach rather than a time-step based simulation. The disadvantage of these methods is that they only take into account the variations of load demands, while the operation of CHP systems also depends on the instantaneous variations in fuel cost and energy policies [97]. A second set of papers is therefore studied in which detailed operational strategies are developed. These present algorithms to find the optimal operation point given an electric and thermal demand at a certain time. For each time step a fitness function is minimised; this method therefore requires substantial computational power if used for a life time analysis.
A model of a cogeneration unit is developed that is a balance between these two types; a time step based analysis that tries to emulate the optimal operation point of a cogeneration unit using simple and intuitive rules of operation.
Optimal sizing over life time: Gamou et al. [98] uses a fuel cell system to model a cogeneration system. The energy demand is simulated using a statistical approach. A hierarchical optimisation technique is used that optimises the unit size taking into account the operational planning using a penalty method. It concludes that taking into account the demand uncertainty in the design phase leads to smaller variation in the annual total cost. In Beihong and Weiding [97], a gas turbine cogeneration system is evaluated on its economic performance. The operational strategy is decided using one representative day for each month. Both the technology and the capacity of the installation are optimised using mixed-integer nonlinear programming. In Ren et al. [99] the model consists of a natural gas fired CHP plant, a thermal storage tank and a back-up boiler. It determines the optimal design capacity of these compon- ents, as well as their optimal operating strategy with the aim of minimizing the annual cost of the energy system. The capacity of CHP plant is taken as an exogenous variable that varies between 0-2kW and the power to heat ratio is assumed to be constant. A sensitivity analysis for the natural gas price, the electricity price and a carbon tax is carried out. It concludes that each project requires its own study to find the optimal in- stalled CHP capacity, and that the results of this analysis is sensitive to the capital costs and energy prices. A thermal storage tank can make the installation more profitable as Chapter 3. Hybrid Power systems 47
long as it is not too large in order to limit heat losses. Ghadimi et al. [91] model a reciprocating engine type CHP unit in combination with an auxiliary boiler and a grid interconnection. It simplifies the operating of the installation by assuming a constant power to heat ratio and no limits on the electric operation point (0%-100%). The evaluation is based on the system’s economics, the primary energy resource consumption and environmental impact. According to the results, in order to optimise the entire energy system it is not sufficient to perform segregated sizing optimisation. Instead, integrated system sizing and operational strategy selection makes sure that the cogeneration unit works together cooperatively with the other system components. Zhang et al. [100] perform a multidimensional life-cycle analysis of four different CHP systems with an additional boiler and thermal storage unit. Three different scenarios are studied. These show the difference between one installed technology compared to multiple, and the possibility of heat dumping compared to no heat dumping. In each scenario the microgrid is connected to the utility grid to meet the electricity demand. It concludes that a microgrid with multiple CHP technologies presents a more economical and environmentally solution to meet the thermal and electric demand.
From this literature we can see that for the sizing process quite a few assumptions are made to simplify the calculations, and instead of simulating an entire year, only a few typical days are analysed. Furthermore, in order to take into account different technologies the time component is reduced to a statistical distribution. This will result in a less accurate system evaluation. Also, short term models are not reliable to determine the optimal system configuration, because more operating conditions during a much longer period should be considered for that purpose [101].
Instantaneous optimal operating point: A more complex way to model and optimize CHP based heating systems is based on a feasible operating region that is represented by a convex area determined by extreme operating points [101]. This model assumes that any point between two of the CHP’s operating points will also be a viable operating point. This is shown in figure 3.17. The marked nodes are the extreme operating points, and the shaded area is the set of all possible operating points. It is important to note that the ratio of installed
thermal power Qinst.CHP and installed electrical power Pinst.CHP depends on the selected technology. Below a certain minimum electrical load the CHP cannot operate and is shut down [101] [102]. In this study this limit is half of the installed power. Chapter 3. Hybrid Power systems 48
(P3,Q3, η3) 100
85 (P4,Q4, η4)
(P2,Q2, η2) inst.CHP % of Q (P6,Q6, η6) 20
(P1,Q1, η1) (P7,Q7, η7) 50 100
% of Pinst.CHP
Figure 3.17: CHP operating region
The cost functions, and power and heat production can then be written as
C = cjxj j∈J X P = pjxj (3.8) j∈J X Q = qjxj j∈J X with xj = 1 j∈J X (3.9) xj 0, j J. > ∈ In this way, constraints can be constructed that take into account the balances for heat, power and storages, system control, power ramping, maintenance, etc. This way of mod- elling is shown in [101], [103], [102] and [104]. Wang et al. [101] use a linear programming (LP) method to minimise the overall net acquisition cost for energy, looking at different CHP systems and storage scenarios. Lahdelma and Hakonen [103] also use LP but solve using a specialised Power Simplex algorithm that efficiently utilises the CHP’s specific matrix structure. In this way the dispatch problem can be solved time-efficiently on an hourly basis over the period of one year. Nazari-Heris et al. [102] compare two differ- ent CHP units by means of a multi-objective cost and emission optimisation using the -constraint method. Demand response, a fuel cell with hydrogen storage and electrical storage and a heat buffer tank are also taken into account. Alipour et al. [104]’s ap- proach is more economic by taking into account an optimal bidding strategy, heat sales to industrial consumers and a demand response program. Non-linear power curves are modelled by Milan et al. [105]. Chapter 3. Hybrid Power systems 49
However, this way of handling optimal dispatch is too demanding in terms of computational power if the calculation is carried out over the entire life time. Therefore a compromise is found between accuracy and speed.
Developing a model: In Nazari-Heris et al. [102], the operational cost function is written as:
i i · i2 · i · i2 · i · i · i C(PCHP,QCHP) = a P CHP+b PCHP+c+d Q CHP+e QCHP+f PCHP QCHP (3.10)
i i with PCHP and QCHP the instantaneous electric and thermal set points of the installa- tion. Example values for the coefficients are shown in table 3.6
Table 3.6: CHP cost function coefficients
a b c d e f 0.0435 56 12.5 0.027 0.6 0.011
200
150
100
operation 50 C
0
1.5 2.5 1 2 1.5 0.5 1 0.5 P [MW] 0 0 P [MW] thermal electric
Figure 3.18: Operational cost of CHP
A two dimensional plot of this function (Figure 3.18) shows a tilted flat surface. This means that the marginal costs for electricity and heat are practically constant across the region of operating points. The marginal cost of heat is very low, around 0.7 e/MWh, which means that extra heat is produced at practically no extra cost. Intuitively this i means that at each set point PCHP of the produced electrical power, a certain amount i i of heat will be generated QCHP = f(PCHP). Either this heat is recovered to meet the thermal demand, or the heat is dumped if the thermal demand is smaller than the produced power; both scenarios will lead to the same fuel cost and cost for operating and maintenance. The set of operating points are therefore limited to the upper boundary Chapter 3. Hybrid Power systems 50
of the shaded region in Figure 3.17. For simplicity, the operating region is further linearised as can be seen in Figure 3.19. The thermal and electrical are also assumed
linear and shown in Table 3.7 [106]. The total efficiency ηtot is the sum of the separate
electric efficiency ηel and the thermal efficiency ηth.
Table 3.7: Cogeneration prime mover efficiencies
Electric load level ηel ηth ηtot 0.5 0.20 0.57 0.77 0.75 0.23 0.545 0.775 1 0.26 0.52 0.78
ηth = 0.52 ηel = 0.26 100
inst.CHP 60 ηth = 0.57 ηel = 0.20 % of Q
50 100
% of Pinst.CHP
Figure 3.19: Linearised CHP operating region
3.3.3 Modes of operation
For CHP installations, three general modes of operation can be identified [101]:
Constant production Depending on the size of the installed thermal power and the system’s annual heat demand, a generation unit can be made to run continuously at full load. This requires that the thermal demand profile be sufficiently season-invariant and the average thermal load be large enough to minimise heat dumping. Furthermore the installation’s elec- trical efficiency and the spark spread have to be taken into account in the sizing of the installation. There are different ways to limit the dumping of excess thermal and elec- trical energy. A bi-annual ON-OFF decision can be made based on the date; ON during Chapter 3. Hybrid Power systems 51
a climatically colder fraction of the year, OFF during the rest. Installing an absorp- tion chiller or thermal storage are other options to optimise heat usage [27].The excess electricity, for example during times of low electric load and high renewable production, can be injected in the nearby utility grid.
Electric load following Under electric load following, the installation is made to meet the system’s electric demand as far as its operating region allows. This mode of operation can be used when the microgrid’s grid connection is not so strong and self-sufficiency is important. Auxiliary gas boilers are required to assure that the instantaneous heat demand is met. When the microgrid is acting in island mode, for example during a grid power supply failure, electric load controlled CHP installations can provide the necessary power to keep the grid stable. Ancillary services to the utility grid are also a possibility when the generator is electrically controlled and a strong grid connection is present.
Thermal load following Under thermal load following, the set point of the CHP installation is determined by the load’s heating demand. Any electricity that is produced in the process is first used to meet the local electrical load, and excesses are injected in the nearby utility grid.
Variations and combinations of these three operating modes can be constructed depending on the characteristics of the thermal and electric load and the prices for electricity and gas. Generally the profitability of CHP installations increases with the yearly operating time. A minimum for economic viability is around 4000 to 5000 hours per year [27]. Mongibello et al. [106] compare two different operation modes, one that allows heat dumping and another that allows load partialisation. It finds that the difference between the two strategies is minimal. However, a hybrid operating strategy that takes into account the instantaneous electricity price and switches between the two modes could present a higher economic performance.
In the section below, two different operation modes are discussed and economically evaluated. Both allow load partialisation and heat dumping. No thermal storage unit is included.
3.3.4 Dynamic techno-economic analysis
The system is modelled using the method described above, with the region of operating points as in Figure 3.19. This way each electric operating point corresponds to exactly one thermal operating point, which simplifies the model significantly. The power to heat ratio in full electric load is taken to be 2, which is a standard value for reciprocating engines. Figure 3.20 and Figure 3.21 show the industrial electric and thermal load profiles used in the simulation. Chapter 3. Hybrid Power systems 52
They are respectively based on the electric consumption of a Belgian industrial park and a generic thermal load profile provided by the Flemish Energy regulator. The total annual electric consumption is 7953 MWh and the annual thermal consumption is 8000MWh.
If the heat generated by the CHP does not meet the thermal demand, an auxiliary boiler is used. If there is excess heat, this heat is dumped. Any surplus or deficit of electricity is compensated using the grid connection. The capacity of both the grid connection and the auxiliary boiler is assumed to be big enough to assure that the thermal and electric load can always be met.
1600 3000
1400 2500
1200 2000
1000 1500
800 1000 Electric load profile (kW) Thermal load profile (kW)
600 500
400 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 hour hour
Figure 3.20: Hourly electric load profile Figure 3.21: Hourly thermal load profile
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 hour
Figure 3.22: Hourly Belpex price
The gas price is taken to be constant at 32.8 e/MWh and includes a fee for distribution. This price is based on the V-test by the flemish energy regulator for a small industrial consumer. The boiler has an efficiency of 80%. Excess electricity is bought from the grid at the instant- Chapter 3. Hybrid Power systems 53 aneous Belpex price plus a fee per MWh for transmission and taxes. The Belpex price time series for 2015 is used, shown in Figure 3.22 and has a mean value of 44.7 e/MWh. The fee for transmission, distribution and taxes is 80 e/MWh; this way the average electricity price is conform with the Eurostat average electricity price for industrial consumers. Electricity is sold to the grid at the Belpex price.
In the simulations for each of the control strategies, the installed electric power is varied between 10% and 100% of the peak electric demand. The installed thermal power is each time twice that amount.
Table 3.8 gives an overview of the simulation parameters
Table 3.8: Overview simulation parameters for cogeneration case
Electric demand: 7953 MWh/year Thermal demand: 8000 MWh/year
Pinst.CHP : 0.156-1.560 MW
Qinst.CHP : 0.312-3.120 MW Electricity price: Belpex (e/MWh) Grid and taxes: 80 e/MWh Gas price: 32.8 e/MWh Life time: 25 year Interest rate: 3 %
3.3.5 System evaluation
Because in the case of cogeneration there is production of both heat and electricity, a COE per MWh (as calculated for wind and solar power) is not as easily defined. For this reason the system will be evaluated using the life cycle cost (LCC), which represents the discounted total cost for covering the load’s electric and heat demands over its lifetime.
LT OPEXf,CHP + OPEXv,CHP + fuel cost + grid cost LCC = CAPEXCHP + (3.11) (1 + r)j j=1 X
The CAPEX depends on the size of the installation and is calculated using the installed electric power Pinst.CHP [91].
2 CAPEXCHP = (8e-5 · P 0.51 · Pinst.CHP + 2e3) · Pinst.CHP (3.12) inst.CHP −
The OPEX consists of a fixed term OPEXf,CHP, and one term that varies OPEXv,CHP with Chapter 3. Hybrid Power systems 54
the annual produced electrical power ECHP [91].
OPEXf,CHP = 1.1e-3 · Pinst.CHP + 6.4152 − (3.13) OPEXv,CHP = ( 8e-7 · Pinst.CHP + 0.0132) · ECHP −
The amount of fuel consumed is a function of the operating points of the CHP and the auxiliary boiler. At any point in time, the fuel consumption for cogeneration is the sum of the produced electric and thermal power divided by the total system efficiency. The boiler fuel consumption is the boiler’s heat generation divided by its efficiency. The boiler’s efficiency is taken to be constant at 80%.
i i i i PCHP + QCHP Qboiler fuel = i i + ηel + ηth ηboiler 8760 (3.14) Annual fuel cost = fueli · gas price i X
The annual grid cost is calculated according to Equation 3.15.
8760 Annual grid cost = grid costi i X (3.15) P i · (Belpexi + fee),P i 0 grid costi = grid grid ≥ P i · Belpexi,P i < 0 grid grid 3.3.6 Reference case
The simulations with cogeneration are compared to a reference case without cogeneration where the entire electric demand is drawn from the grid and the entire thermal demand is generated with a gas boiler. Economic profitability demands that the total energy cost with cogeneration be less than in the reference case.
The total energy cost in the reference case is 23.218 million e.
3.3.7 Electric load following:
Electric load following is implemented as described in subsection 3.3.3. The decision algorithm i i is presented in Figure 3.23. LPE and LPT are respectively the electric and thermal demand i i and PCHP and QCHP the electric and thermal operating point at time step i. Chapter 3. Hybrid Power systems 55
i i LPE, LPT
i i PCHP = QCHP = 0 yes i no i i ON LPE PI/O OFF Pgrid = LPE ≥ i i Qboiler = LPT
i PCHP = Pinst i i no i i Pinst LPE Pgrid = LPE Pinst ≥ i −i QCHP = f(Pinst)
yes i i PCHP = LPE i i i no i Pgrid = 0 QCHP < LPT Qboiler = 0 i i QCHP = f(PCHP)
yes
Qi = LP i Qi boiler T − CHP Figure 3.23: Electric load following control algorithm
For each time step the electric demand is compared to a reference value PI/O, that determines the ON-OFF behaviour of the installation. It can be written as a fraction of the installed electrical power. If the electric load is lower than the decision variable, the CHP is turned off and all electric and thermal power is supplied using the grid and the auxiliary boiler. If the load profile is higher than the decision variable, the cogeneration unit is switched on and operated to meet the electric load as far as its operating region allows. For electric loads higher than the installed electric power, the grid is used to import the additionally required power. The thermal set point is an unambiguous function of the electric set point, as shown in Figure 3.19. As a final step, the thermal generation is compared to the thermal demand and any additionally required heat is generated with the auxiliary boiler. Surplus heat is dumped.
Two variables are varied. The installed electric power Pinst.CHP is varied between 10% and
100% of the load’s peak demand. The decision variable PI/O is varied between 50% and 100% of the installed power. For each case the total energy cost is calculated. These results are Chapter 3. Hybrid Power systems 56 presented three-dimensionally in Figure 3.24.
The optimal size for the cogeneration unit lies around 40% of the peak demand, which in this case is about 624 kWe and 1248kWth. The LCC is 20.858 millione, which is below the cost for the reference case. The decision set point for this installation size is not of importance; because of the season-invariance of the electric load profile (see Figure 3.20) all set points are below the minimum power demand and the CHP unit is switched on throughout the year. For higher installed powers, the total energy cost increases with an increasing value of the decision value. This confirms that a higher usage of the installation increases its competitiveness.
×107
2.6
2.5
2.4
2.3 100% 90% 2.2 80% 2.1 70% 60%
2 Decision set point 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Installed electrical power
Figure 3.24: Simulation results with electric load following
3.3.8 Thermal load following:
The algorithm for thermal load following is shown in Figure 3.25. It is analogous to electric load following, except that the IO-decision is now made according to the thermal demand and the thermal installed power. Also, the electric set-point is decided in function of the thermal set point, and not vice versa.
The optimal installed electric power is again 624 kWe (1248 kWth), however under this operating strategy the LCC is 21.386 million e, which is below the reference case but higher than the cost for electric load following. The decision variable is important even at low amount of installed power, as can be seen by the rising cost for higher decision set points. This is because the thermal load profile is much more season-dependent (see Figure 3.21). Chapter 3. Hybrid Power systems 57
i i LPE, LPT
i i QCHP = PCHP = 0 yes i no i i ON LPT QI/O OFF Qboiler = LPT ≥ i i Pgrid = LPE
i QCHP = Qinst i i no i i LPT Qinst Qboiler = LPT Qinst ≤ i −i PCHP = Pinst
yes i i QCHP = LPT i i i Pboiler = 0 Pgrid = LPE PCHP i i − PCHP = f(QCHP)
Figure 3.25: Thermal load following control algorithm
×107
2.6
2.5
2.4 100%
2.3 90% 80% 2.2 70%
2.1 60% Decision set point 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Installed electrical power
Figure 3.26: Simulation results with thermal load following Chapter 3. Hybrid Power systems 58
3.3.9 Conclusion on CHP model
The above results suggest that a control algorithm based on the combination of load partial- isation and heat dumping can make a CHP installation more economical than separate heat and power production. Both electric and thermal load following result in the same amount of installed electrical capacity, which is about 40% of the peak. Of the two algorithms, electric load following results in a lower total cost, and will therefore be used in the rest of this study. The decision set point is taken to be the physical lower boundary of the installations operating region, which is half of the installed capacity.
3.4 Electrical storage
Due to the radical increase of intermittent generation units, electrical storage might prove to be crucial in the future’s electrical power grid. It can can play a set of different roles such as help manage future peak demands, allow dynamic energy management, aid the introduction of renewable power units and improve power quality [107]. Depending on the nature of the application, a different technology can be installed. Table 3.9 gives an overview of the technical characteristics of different types of electrical energy storage (EES). Three main categories can be identified: mechanical storage, electrical storage and chemical storage. Mechanical storage can store electrical energy in two ways, either through the conversion to gravitational potential energy or kinetic energy. Electrical energy storage relies on electrostatic or electromagnetic fields to store the energy. Chemical storage uses electrochemical conversion for energy storage.
While electrical storage can provide crucial services to the utility grid, its high cost is a severe obstacle for a large scale roll out. Besides the potential of electrical storage for ancillary services to the grid, long term storage and seasonal storage can mitigate the effect of periods with low wind and low solar radiation on a power system with high integration of renewables. In an industrial context, storage can be economical for protection against power drops and improving power quality, but also for lowering the electricity cost through load shifting. Furthermore, storage can be part of a UPS system instead of the more conventional diesel generator [29].
In microgrids, energy storage is vital if island mode is considered and a pre-requisite for good power quality [108]. Table 3.1 shows that all but one of the studied papers include electrical storage in their hybrid power system. Even for grid-connected hybrid systems that can use the grid as a system back-up, electrical storage can be used to enhance self-consumption and reduce grid interactions to those that are economically lucrative. Chapter 3. Hybrid Power systems 59
Table 3.9: Technical parameters for different storage technologies [29]
Round trip Depth of Lifetime in Self-discharge efficiency in discharge in Deployment years per hour in % % % time Now 2030 Now 2030 Now 2030 Now 2030 Mechanical: Pumped hydro 75-82 80 80-100 5e-4 3 min Compressed air 60-70 25 35-50 3e-2 5 min Flywheel 80-95 15 75 10 10 ms Electrical: Super-capacitors 90-94 15 75 0.52 < 10 ms Superconductive 80-90 20 na 0.5 magnetic Chemical: Hydrogen 34-40 40-50 40-60 6.9e-4 10 min Flow batteries 60-70 65-80 10-15 15-25 100 1.1e-2 5.2e-3 1-2 s Lithium-ion 83-86 85-92 5-20 10-30 100 6.9e-3 1.4e-3 3-5 ms Lead acid 75-80 1 5-15 10-20 70 80 1.1e-2 5.2e-3 3-5 ms High Temperat- 75-80 80-90 5-15 20-30 100 0.42 3-5 ms ure Batteries
3.4.1 Modelling
The battery is modelled as a storage device that fulfils three functions: to charge, to hold and to discharge energy [109]. Unlike in the case of cogeneration, there is little variation in the way electrical storage is modelled in literature. The studied papers all start from a perfect storage device with limited storage capacity and introduce different system constraints and imperfections to derive a realistic storage model. Practically all of the studied papers use the characteristics of a lead acid battery as a starting point for their analysis.
The important variables that characterise the battery model are the following. Einst.bat is the energy capacity of the system and is the maximum amount of energy that can be stored in the system. The state of charge SOCi is how much energy the storage contains at the end
of time step i as a fraction of its maximum Einst.bat. The exchanged power with the storage i i system P exbat is how much power is either fed to the EES (P exbat > 0) or how much power i is drawn from the EES (P exbat < 0) per time step. The model assumes that all power is injected or withdrawn at the same time; at the end of the time step. Chapter 3. Hybrid Power systems 60
Table 3.10: Economic parameters for storage technologies
Power cost (e/kW) Energy cost (e/kWh) Now 2030 Now 2030 Mechanical: Pumped hydro 500-1000 5-20 Compressed air 1000 700 40-80 Flywheel 300 1000 Electrical: Super-capacitors 10-20 10000-20000 Chemical: Hydrogen 1500-2000 500-800 0.3-0.6 Flow batteries 1000-1500 600-1000 300-500 70-150 Lithium ion 150-200 35-65 300-800 150-300 Lead acid 150-200 35-65 100-250 50-80 High Temperat- 150-200 35-65 500-700 80-150 ure Batteries
In the ideal situation when no constraints are taken into consideration, the relation between the aforementioned variables is: P exi SOCi = SOCi−1 + bat (3.16) Einst.bat
Four different types of constraints can be distinguished in literature .
Power constraint: Under the power constraint, the power drawn or stored in the battery during one time
step cannot exceed a maximum value Pinst.bat. Thus for every time step the absolute i value of the power exchange with the battery P exbat should be within a certain limit.
i 0 < P ex < Pinst.bat (3.17) | bat| This constraint is used in [74], [33] and [83], amongst others. In many cases the power constraint is part of the set of system parameters that is optimised as part of the
optimal sizing. In that way it is treated just like the installed storage capacity Einst.bat.
An increase in EES power Pinst.bat is possible but will come at a certain cost.
Efficiency constraint: The efficiency constraint takes into account the power loss that occurs when energy is converted in the storage system. This constraint is used in practically all of the studied Chapter 3. Hybrid Power systems 61
articles and for large storage systems it will be a source of significant losses. It concretely i means that in the case of charging (P exbat > 0), the power fed to the battery is greater than the energy that is effectively stored. Likewise, when a certain amount of energy i P exbat (< 0) is withdrawn from the system, only a certain fraction of this energy will be delivered to the grid.
i i−1 P exbat i SOC + ηcharge · charging (P ex < 0) i Einst.bat bat SOC = P exi (3.18) SOCi−1 + 1 · bat discharging (P exi > 0) ηdischarge Einst.bat bat In theory the charging efficiency ηcharge and the discharging efficiency ηdischarge will have different values. However, separate values for these parameters are hard to measure [76] and battery manufacturers will therefore define a round trip efficiency for one entire charge and discharge cycle. Round trip efficiency values for different storage systems are shown in Table 3.9. In the model used here the charge and discharge efficiency are
assumed to be equal, ηcharge = ηdischarge = ηbat. ηbat is taken to be equal to the square root of the round trip efficiency.
ηbat = √ηroundtrip (3.19)
Some authors such as [77] and [73] assess the inverter losses separately. Here these will
be included in the system parameter ηbat.
Self-discharge: Storage systems exhibit self-discharge behaviour, which means that over time a certain fraction of the stored power will be dissipated and lost [110]. The amount of self- discharge can be expressed in percent per hour; values for different technologies are shown in Table 3.9. It is denoted with the symbol σ. In the battery model it means that independent of the exchanged power, the SOCi at a certain time is less than the SOCi−1 of the time step before.
SOCi = (1 σ) · SOCi−1 (3.20) − Self-discharge is taken into account in [72], [73], [76], [80], [83] and [84].
Depth of discharge and maximum SOC: By introducing a maximum depth of discharge of the EES the longevity of the system can be prolonged [110] [111]. Concretely this means that when the battery is at its minimum state of charge, it no longer allows power to be withdrawn. Different technologies have different values for the minimum state of charge. As a constraint it is possible to write
i SOCmin < SOC < SOCmax (3.21)
SOCmax is in practice almost always equal to 1 and typical values for SOCmin are between 30% and 0%. Fossati et al. [74] identifies that this limit does not take into Chapter 3. Hybrid Power systems 62
account the possible trade-off that can be made between reduced life time and instant-
aneous benefit of working below the SOCmin. However, aging will not be taken into account in this work, so this trade-off cannot be evaluated.
Besides these constraints, calculating the expected lifetime of batteries is important since it can significantly influence the cost of the system [78]. Different methods can be used to determine the lifetime of the system, most of which count the cycles to failure. Furthermore, since electrical storage is not yet a mature technology, big differences can be expected in the technical as well as the economic parameters of storage systems between now and in 10 years (see Table 3.9). While generating units’ expected lifetime is at least 20 years [71], this is clearly not the case for current electrical storage technology. This model is based on a lead acid battery and the expected life time is 12.5 years. Cost estimations for batteries now and in ten years are shown in Table 3.10.
3.4.2 Battery control
Two different battery control systems were implemented. The first algorithm only takes into account the power production and consumption in the grid at each time step and tries to flatten the interaction with the grid by maximally using the battery’s capacity. The second algorithm is a fuzzy logic based control structure that takes into account the instantaneous electricity price as well as the battery’s technical constraints.
Control I - maximal usage
The first battery control strategy is based on maximally using the battery storage to limit interaction between the microgrid and the utility grid. This battery control strategy is used by Koutroulis et al. [71], Li et al. [72], Tafreshi et al. [73] , Abbes et al. [77], Tito et al. [80] and Theo et al. [83]. Figure 3.27 gives an overview of the algorithm. The input parameters are the previous state of charge SOCi−1 and the instantaneous power flow at time i, PF i. The power flow is the instantaneous difference between the electric production and the electric consumption in the microgrid at time i. This is the power that would normally be drawn from or injected in the grid. A control decision is made solely based on whether this power flow is positive or negative. If PF i > 0, there is a netto excess of electricity in the grid and power can be stored in the battery; likewise, for PF i < 0, there is a deficit of electricity that will require the battery to discharge. In both cases the magnitude
of the power flow is compared to the maximum power Pinst.bat that can be drawn from Chapter 3. Hybrid Power systems 63
the battery. Any power above this value can not be compensated by the battery and will need to be exchanged with the grid. The grid interaction because of the battery’s
power limitation is called Pgrid1 and corresponds to the first constraint discussed above. The resulting formulas for both charging and discharging are shown below. Charging Discharging
i i if: PF > P if: PF > Pinst.bat inst.bat | | i i i i Pgrid1 = PF Pinst.bat Pgrid1 = PF + ηbatPinst.bat − (3.22) (3.23) i i PF = Pinst.bat PF = Pinst.bat − The second step in the control algorithm checks to what extend the storage system’s previous state of charge allows the power transfer of PF i. This step takes into account the second, third and fourth constraints discussed above. In the case of a positive PF i (charging algorithm) the amount of vacant storage capacity is checked and filled accordingly. This can be written as:
i−1 Einst.bat i if: (SOCmax (1 σ)SOC ) · < P F − − ηbat i SOC = SOCmax (3.24) i i−1 · Einst.bat Pgrid2 = PFi (SOCmax (1 σ)SOC ) − − − ηbat
else: η PF i SOCi = (1 σ)SOCi−1 + bat − Einst.bat (3.25) i Pgrid2 = 0
For a negative PF i (discharging algorithm) the formulas are analogous. Instead of vacant capacity the constraint depends on the amount of stored capacity. i−1 i if: ((1 σ)SOC SOCmin) · ηbatEinst.bat < PF − − | | i SOC = SOCmin (3.26) i i i−1 P = PF + ((1 σ)SOC SOCmin) · ηbatEinst.bat grid2 − − else: PF i SOCi = (1 σ)SOCi−1 + − ηbatEinst.bat (3.27) i Pgrid2 = 0
The value for the total grid exchange in both charging and discharging is calculated as
Pgrid = Pgrid1 + Pgrid2. Chapter 3. Hybrid Power systems 64
SOCi−1, PF i
yes no Charge PF i > 0 Discharge
i Power i PF < Pbat PF < Pbat constraint | |
SOC SOC < SOCmax SOCmin < SOC constraint
i SOC , Pgrid
Figure 3.27: Battery control algorithm - maximal usage
While this algorithm is able to do a fair job at modeling the technical battery behaviour, it is not very effective as a battery dispatch algorithm in a grid-connected microgrid. For example, when the microgrid’s own production is not so big and it depends on some constant amount of grid power for its day-to-day operation, the value of PF will be negative for every time step and the battery will never charge. Furthermore this algorithm is oblivious to economic incentives, which is important in making electrical storage cost effective. It is important to note that the articles in which this strategy is implemented mainly treat stand-alone systems in which there is no grid connection. As such the battery must be used maximally in order to assure electric supply as much as possible. In a grid-connected microgrid the storage system’s objective has to be more complex. Otherwise batteries can be omitted from the hybrid system; Grid-connected systems do not necessarily require battery banks, and they mean as much as 50% of the investment cost [81]. A smart control strategy should be implemented that takes into account the variations in the grid’s electricity price. Chapter 3. Hybrid Power systems 65
Control II - fuzzy logic
For the reasons mentioned above, a fuzzy logic based control algorithm is developed that takes into account the instantaneous electricity price in the control of the battery storage. It is a simple version of the system that was developed by Fossati et al. [112] that was later used in a publication on optimal sizing by the same authors [74]. The controller’s input is the instantaneous electricity price and the power flow in the
grid. From these it calculates an output Pfuzz, which is the battery’s economic set point for that time step. This value is fed to an algorithm that compares the economic set point to the battery’s physical state and it determines what power exchange can take place.
pricei, PF i
Fuzzy Control
i Charge Pfuzz Discharge
Power
i constraint i P < Pbat P < Pbat fuzz | fuzz| SOCi−1
SOC SOCi < SOC SOC < SOCi max constraint min
i SOC , Pgrid
Figure 3.28: Battery control algorithm - fuzzy logic
The strength of a fuzzy logic based control structure is that it allows to simplify decision processes that are based on multiple inputs. It does this by emulating the vagueness of a human cognitive process and thereby allows to take decisions under uncertainty [111]. Chapter 3. Hybrid Power systems 66
Fuzzy logic steps away from the crisp set theory in which an object (a number) is either a member or not a member of a certain set. Instead, it uses membership functions to define to what extend an object is part of a set [113]. This way of reasoning is intuitive because the fuzzy sets can be defined according to linguistic variables such as ”High”, ”Low”, ”Good”, ”Bad”. The output of the membership functions is a fuzzification of the original input variable. Defining a control structure with these qualitative terms is done according to linguistic rules and is therefore simple. An example of a rule is:
If Input A is ”NOT Low” and Input B is ”Very High” then Output is ”Negative”.
The interpretation of this is that the output variable belongs to the fuzzy set ”Negative” up to the point that the stated premise is true. In other words, the output is a member of the subset ”Negative” to the extend that Input A is not a member of the subset”Low” and Input B is a member of the subset ”Very High”. The combination of all these rules for a set of input variables results in a set of fuzzified output variables. The aggregated output is translated to crisp values using output membership functions [114]. In the controller developed here, the input variables are the instantaneous electricity selling price pricei and the power flow in the microgrid PF i. They are normalised using the following formulas. i i PF PFmin PFnorm = − (3.28) PFmax PFmin − pricei price pricei = − min (3.29) norm price price max − min pricemax and pricemin are respectively the highest and lowest electricity price of the year. PFmax and PFmin are the yearly highest and lowest power flow in the grid The fuzzification membership functions for the normalised inputs are shown in Fig- ure 3.29 and Figure 3.30. They are trapezoidal and the categories are ”Low” (L), ”Medium” (M), ”High” (H), Very High ”VH”. For PF there is also a membership func- tion ”Very Low” (VL). The specific membership functions are constructed taking into account the characteristics of the normalised variables. For the electricity price, the mean of the normalised hourly Belpex price lies around around 0.1, which is the ab- scissa of the peak of the ”Medium” function. Likewise, ”VH” is chosen such that only 1% of the values during the year fall within its range. The normalised power flow is a variable that tends to be more stochastic than the electricity price, which is why the membership functions are more equally space. The linguistic control rules are displayed in Table 3.11. The main aim of this control structure is to charge the battery when the electricity price is low, and discharge when the electricity price is high. This means that the battery can charge even though Chapter 3. Hybrid Power systems 67
VL L M H VH 1
0.5 Fuzzified output
0 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 − − − − − Crisp input
Figure 3.29: Membership functions for power flow
L M H VH 1
0.5 Fuzzified output
0 0 0.2 0.4 0.6 0.8 1 Crisp input
Figure 3.30: Membership functions for electricity price
the power flow in the grid is negative, or discharge when the power flow is positive. However, under no circumstances should the installed battery require fortification of the grid connection (and thus a greater grid dependency). The rules are constructed through trial and error. The basis of the control are the first i 4 rules, which solely look at the electricity price to determine Pfuzz. The final three rules make sure that for very high or very low power flow in the grid, the economic factor plays a lesser important role and technical considerations take over. This is to avoid overloading the microgrid’s grid connection. The membership functions for defuzzification are shown in Figure 3.31. The normalised
value will be multiplied by the battery power capacity Pinst.bat which is why the the range of the normalised variable is [-1 1] instead of [0 1]. Chapter 3. Hybrid Power systems 68
Table 3.11: Fuzzy control rules for battery storage
i i i price PF Pfuzz L / VP M / Z H / VN VH / VN (L) VH N ¬ (VH) VL P ¬ (H) VL P ¬
VN N Z P VP 1
0.5 Crisp output
0 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 − − − − − Fuzzified input
Figure 3.31: Membership functions for Pfuzz
3.5 Demand-side management
Demand-side management in the context of power systems entails all the measures that have an impact on the consumption side of an energy system. While the idea of demand management is definitely not new, the current trend of distributed, volatile generation and economical communication techniques have created an environment in which dynamically adapting the consumption to the production is less expensive than vice versa [115].
Three different categories of DSM can be identified:
Energy Efficiency Improving the energy efficiency of a power systems’ loads results in a permanent decrease in energy consumption.
Demand Response (DR) Demand Response programs shift or reduce electrical consumption, either directly or according to economic incentives, in order to flatten or reduce the demand profile [116]. Chapter 3. Hybrid Power systems 69
Demand response programs can be used at critical times and has the ability to both increase power grid reliability and potentially reduce operating system costs [117]. An example of a DR program is Time of Use (TOU) electricity tariffs in which a supplier charges different rates for electric consumption at different times of the day. This allows utilities to use financial incentives to convince costumers to shift their consumption from times of high demand to times of low demand.
Spinning Reserve Spinning reserve by demand management supports the utility grid by load curtailment in critical moments. It could be considered a type of demand response. According to Kirby [118], spinning reserve through demand management is the most underutilised reliability resource available to the power system today .
In this study the potential of demand response in a microgrid will be assessed. Three different DR actions can be identified.
- Load shedding
- Load shifting to later time
- Load shifting to earlier time
Gils [119] make an assessment of demand response potential in Europe. For the belgian tertiary sector this results in a total estimate of 6337 MW of DR potential. The annual tertiary energy consumption in Belgium is 22.18TWh. This results in a DR potential of 10.4% of the average power demand.
3.5.1 Demand response modeling
Demand response is often modelled as a economic mechanism that depends on the price elasticity [31]. In this study a heuristic approach is used that looks at the daily demand profile and shifts a certain percentage of the total daily energy consumption from times of high to times of low consumption. Examples for the original and resulting power flow per day and per week are shown in Figure 3.32 and Figure 3.33.
Figures 3.34 and 3.35 show the annual instantaneous interaction with the grid in the case with and without demand response. It is clear that with DR the intra day variation is reduced. Table 3.12 presents the reduction in annual peak power in function of the percentage of daily shifted energy consumption. It shows that a peak reduction of around 10% corresponds to shifting 2.5 % of the load. Chapter 3. Hybrid Power systems 70
0 500 no DR without DR -200 DR 1% with DR DR 5% 0 -400
-600 -500
-800 -1000 -1000
Grid power in kW -1200 Grid power in kW -1500
-1400 -2000 -1600
-1800 -2500 2 4 6 8 10 12 14 16 18 20 22 24 0 20 40 60 80 100 120 140 160 hour hour
Figure 3.32: Demand response over one day Figure 3.33: Demand response over one week
2000 2000
1000 1000
0 0
-1000 -1000
-2000 -2000 Grid power with DR (kW) Grid power without DR (kW)
-3000 -3000
-4000 -4000 0 1000 2000 3000 4000 5000 6000 7000 8000 0 1000 2000 3000 4000 5000 6000 7000 8000 hour hour
Figure 3.34: Pgrid without DR (entire year) Figure 3.35: Pgrid with DR (entire year)
Table 3.12: Demand response simulation results
Daily shiftable load Peak load Peak reduction 0% 100% 0% 0.5% 95.3% 4.7% 1% 92.9% 7.1% 2.5% 89.3% 10.7% 5% 84.5% 15.5% Chapter 4
Optimal design and sizing
4.1 Evaluation of electricity production system
The model as described in Chapter 3 determines the dynamic behaviour of a microgrid ac- cording to its technical characteristics such as the load profile, installed wind power, solar power, etc. In order to optimise this system, evaluation functions have to be defined that evaluate the system’s performance.
Concretely an evaluation function f will be of the form
f(x1, x2, ..., xm) = system evaluation (4.1) with x1, x2, ..., xm etc. the technical specifications of the installation.
Table 4.1 gives an overview of the different evaluation functions and constraints found in literature. These are explained in the section below. The abbreviations are explained in Table 4.2.
4.1.1 Economic evaluation
Life Cycle Cost (LCC): The Life Cycle Cost is the most used evaluation method in the studied literature. It is an indication of the total system cost over its lifetime, and takes into account the market’s general interest rate. The annual costs are discounted and added according to equation 4.2 [81] [71] [83].
LT Cj LCC = C + annual (4.2) initial (1 + r)j j=1 X
LT is the lifetime of the installation, Cinitial is the initial investment cost, Cannual the annual cost and r the interest rate. The initial investment cost depends on the size of the installed technologies and is shown
71 Chapter 4. Optimal design and sizing 72
Table 4.1: Overview of optimisation methods in studied literature
Evaluation Optimisation Constraints Load function algorithm Koutroulis et al. [71] LCC GA LPSP Residential Tito et al. [80] LCC GA LPSP Gonz´alezet al. [81] LCC GA PB Entire township Li et al. [72] LCC GA & homer PB Island Industrial and Theo et al. [83] LCC MILP PB residential Li et al. [85] LCC Novel Tafreshi et al. [73] COE GA LPSP Remote small Ismail et al. [75] COE GA LPSP communities Yang et al. [86] COE Linear LPSP Fossati et al. [74] MOC GA PB Kamjoo et al. [84] ROI not specified PB Telecommunication Yang et al. [76] ACS GA LPSP station Bayod-R´ujula et al. System none PB [82] losses Residential, Zhou et al. [32] Net profits GA Price elasticity commercial and industrial - LCC Dongfushan Zhao et al. [33] - REP GA PB Island - GHG - LCC Abbes et al. [77] - EE GA Residential home - LPSP - LCC Dufo-L´opez and - GHG 2-layer GA PB Bernal-Agust´ın[78] - LPSP - COE - LPSP Katsigiannis et al. [79] - GHG GA - Fuel cons. - REP Chapter 4. Optimal design and sizing 73
Table 4.2: List of abbreviations and occurrence
abbrev. Evaluation function/Constraint # LCC Life cycle cost 9 COE Cost of electricity 4 ACS Annualised cost of system 1 MOC Mean operation cost 1 ROI Return on investement 1 LPSP Loss of power supply probability 9 PB Power balance 8 GHG Green house gas emissions 3 REP Renewable energy penetration 2 EE Embodied energy 1
in Equation 4.3.
Cinitial = CAPEXsolar · Pinst.solar + CAPEXwind · Pinst.wind
+2 · CAPEXc,bat · Einst.bat + CAPEXp,bat · Pinst.bat (4.3)