I. INTRODUCTION

Transcription

1 GENETIC ALGORITHM FOR OPTIMAL BATTERY ENERGY STORAGE SYSTEMS CAPACITY AND SITE SELECTION FOR A DISTRIBUTION NETWORK WITH A HIGH PENETRATION OF WIND ENERGY 1 ABDULWAHAB ALHAMALI, 2 MOHAMED EMAD FARRAG, 3 GERAINT BEVAN, 4 DONALD M HEPBURN 1,2,3,4 Glasgow Caledonian University, Glasgow, UK 1 Abdul.AlHamali@gcu.ac.uk, 2 Mohamed.Farrag@gcu.ac.uk, 3 Geraint.Bevan@gcu.ac.uk, 4 D.M.Hepburn@gcu.ac.uk Abstract - Distributed generation (DG) sources, e.g. wind turbines (WT) and photovoltaics (PV) are becoming more common in distribution network systems (DNS) and their presence can affect DNS performance. To improve DNS reliability and performance, incorporating energy storage systems (ESS) is becoming increasingly important. ESS can help facilitate the incorporation of DG sources and to resolve technical challenges, e.g. transient stability. Power generated from DG sources is variable, particularly in the case of WT, and as power variability affects the DNS stability, ESS have been used to regulate the demand from the supply network. Given the high cost of mass battery energy storage systems (BESS), finding a method to determine the optimum size of BESS for a given DNS would assist operators. In DNS, voltage and power losses of the network are considered to be significant matters. BESS location affects system operation and power flow within the network, so there is a clear need to find the optimal BESS location in DNS. In this paper, a methodology for optimal siting and sizing of a BESS in DNS with a high integration of energy from WT is presented. The proposed technique is based on genetic algorithms (GA) with power flow (PF) to identify the best placement and capacity in order to achieve optimal DNS performance. The general principle of operation is to charge the BESS through unused energy or off-peak duration and discharge during peak time or low energy generation. The suggested method is verified on a typical UK DNS to validate the performance and effectiveness of the technique. Keywords - Distributed generation (DG), Distribution systems, battery energy storage system, optimal allocation and Genetic algorithms (GA). I. INTRODUCTION Concerns over global environmental changes and growing energy needs have led to greater use of renewable energy devices in electrical distribution network systems (DNS). In such cases DNS operation will be affected, e.g. its voltage stability and serious concerns have been raised over reliable and satisfactory system operation [1]. One proposed solution to increase system reliability and performance is to incorporate energy storage devices into DNS [2]. Energy Storage Systems (ESS) use devices, such as batteries, to facilitate the use of renewable energy sources within DNS and help resolve technical challenges. Photovoltaic (PV) and wind turbines (WT) are the most common renewable energy sources used to produce electricity [3]. Prediction of power generated from WTs is challenging because wind speed and direction are variable and because of the poor long term prediction of local wind conditions. ESS have been used to compensate a local area supply for intermittent renewable sources [4]. A promising area is to embed battery energy storage systems (BESS) with DG units, they can be charged during times when energy prices are low (off-peak) and discharged when energy prices are high (peak). Owners/end-users can benefit from energy price fluctuations and help develop a more environmentally friendly electricity network [5, 6]. At present the cost of BESS is still relatively high [7]. In DNS, voltage profile and network power losses are considered to be significant matters and it has been shown that BESS location can have a positive effect on system operation. As such, there is a clear need to determine optimum size and site for BESS in any network, for example, to consider whether increasing the number of BESSs and arranging for them to work in a coordinated manner would reduce the overall cost and decrease network losses [8]. In order to find the optimum sites for BESS in a specified electrical microgrid (MG), [9] suggested a novel method which relied on a harmony search algorithm to minimize MG cost. This method aims to minimize the amount of energy which would need to be bought from the main electrical grid, however, it only addressed ESS optimal location. [10, 11] proposed novel algorithms for distributing ESS units in DNS (called DESS) to improve operation and reliability and to defer system upgrades: the cost of DESS investment is not considered. In [12], GA has been used to find optimal size and location of DESSs to defer network upgrade and minimizing power losses. This model has been enhanced by [13] to include state of charge (SoC) limitations and the efficiency of charge/discharge cases. In [14], a combination of GA and dynamic 82

2 programming was used to find best site and size of DESSs to reduce power loss and total investment. In [15], Analytical Methods (AM) were used to investigate balance in production from renewable energy sources but applied only for optimally determining ESS capacity. Yao et al. [16] researched determination of both BESS power rating and minimal energy rating (number of battery cells), the former relied on statistical wind distribution, while the latter had to satisfy additional battery specific limitations. To find allowable generation from renewable sources under network constraints, [17] applied a Mixed Integer LP (MILP) and then an optimal power flow (OPF) method to determine the ESS capacity and site to minimize electricity cost and avoid generation curtailment. 2.1 Optimal site and size of BESS: The methodology proposed in this work is based on using genetic algorithm and aims to reduce power losses and improve voltage profile. The GA flow chart used to find the optimal allocation of BESS is presented in Figure 1 and the objective function is outlined below. The objective function: Heuristic methods can be applied to find the optimum ESS allocation in a DNS. This technique has been used in a model of a practical grid to determine the optimal place(s) and size(s) of the BESS [18]. To determine the optimal site and capacity of a BESS in a DNS Genetic Algorithms (GA) can be used. PF is integrated for each solution to optimise the charge/discharge duration of batteries in order to reduce the power losses and improve voltage profile in the DNS. Then the fitness function is evaluated and, so, the resulting solutions are characterized. To determine the best fitness function and reach optimal final solution, several operations have been developed using the population, e.g. reproduction, crossover, and mutation as shown in the Figure 1. In this paper, a GA and OPF methodology is suggested for optimal allocation of DG and BESSs in DNS with a high wind energy penetration. Simulation will be separated into two time periods. Batteries charge through off-peak periods, during peak period batteries can be discharged. BESS capacity will be determined and its site will be optimised to improve voltage profile and decrease network losses. Power flow Constraints The annual predicted wind energy, the load and the battery power are all integrated into OPF which is used to obtain the optimal site and location of batteries to avoid violation of system limitations. II. PROBLEM FORMULATION This work is based on [19], where high penetration of wind energy is determined to be 142% based on 186 WT capable of generating 11kW: [19] did not use energy storage. In this work, BESS will be used with WT to supply the loads with 100 % from renewable energy and storage for the whole year. The percentage penetration of WT will be decreased to meet the minimum number required to satisfy the winter season, as it has the highest load. MATLAB scripts have been used to calculate percentage based on Equation 1. BESS Efficiency = discharged energy /charged energy.. (1) Figure 1: Flow chart of the GA method 83

3 The seven optimization constraints applied are as follows: kw WT has been used to calculate the generated Power. Figure 3 presents the variation of output power for the four seasons [19]. Figure 3: Power output of 11kW Wind Turbine for the four seasons 2.4 Modelling of Storage System In this work the batteries are used to time-shift electric energy from WT. At off-peak hours batteries are charged, when wind power produces more than the load requires, and at peak hours batteries are discharged. Based on Equations 10 and 11, the capacity of storage energy can be calculated [22]. For charging: (10) For discharging.... (11) : Energy stored in the BESS at time : Self-discharge rate of the BESS in one hour. 2.2 Load modelling The load used is an hourly power demand curve, produced from historic data for the UK power network. Figure 2 shows the daily load profile for four seasons [21] (Load factor: Winter =1, Summer = 0.5, Spring & Autumn = 0.75). III. RESULTS AND DISCUSSION Figure 4: LV distribution network in the UK [21] Figure 2: Average Daily load in the UK distribution network From Figure 2, it is clear that electricity consumption in the UK has a regular pattern with high demand between 08:00-22: Wind power generation model For each of the four seasons, a typical day is produced to illustrate the random behavior of the wind speed. Hourly mean wind speed data was obtained from Met Office Archives for a weather station at Bishopton, near Glasgow, between 2011 and Based on this, curve fitting of a typical 11 In this paper, the single line diagram of the LV distribution network in the UK is used to validate the proposed method. Figure 4 shows a substation of a LV distribution network, it has an 11/0.4 kv transformer, rated at 0.5 MVA, with 384 homes being fed [21]. 3.1 Determining the size and location of DGs Based on the previous research [19], the authors developed the algorithm to calculate the maximum number of WT to cover the base load, and simultaneously eliminating reverse power through the distribution transformer, therefore, no need to modify the transformer protection system. The maximum 84

4 number of WTs was 186 units. In this study the author is working on a target to reduce the WTs to a minimum that will satisfy the condition of reverse power and reducing the unused wind energy during low load by utilising the ESSs. Based on the total annually unutilised wind energy, an algorithm is developed to include the ESSs for the purpose to reduce WTs number. As can be seen in Figure 5, the optimum number was found to be 85% of the previous calculation that guaranteed 100% of the load demand is covered by the wind energy generation. The ESSs is considered as bulk storage located at the transformer secondary side. The developed GA has now applied to redistribute the total WTs over the distribution network, providing the constraint of WTs is less than or equal number of homes at the same bus. This result is shown Figure 6. Figure 7: Typical day of each season showing unutilised wind energy and energy needed from the grid (159 DGS) Figure 5: Typical day for each season displaying comparison of presence of 186 and 159 WT on tested network Figure 8: Typical day of four seasons showing unutilised wind power (used in charging case) Figure 6: Tested LV network with distributed DG using GA Identifying unused wind energy and grid demand: The difference between load demand and the generated wind energy has been used to calculate the unutilised wind energy and the energy demand from the grid; hence the capacity of the ESSs could be accurately identified and redistributed across the DN. This also shows the time frame for charging and discharging the ESSs as shown in Figure 7. It is clear that the peak of unutilised wind energy is occurring in spring. However the maximum grid demands as expected to occur in the winter. Figure 8 shows the hourly basis of the unutilised wind energy of over 300 kw in spring. Figure 9 shows the peak grid energy demand of 162 kw is occurring in winter. Figure 9: Typical day of four season showing power needed from the grid. (used in discharging case) 3.3. Determining optimal size and location of ESSs using GA: An OPF and GA have been used to identify an optimal site and size of ESSs. In the solution of OPF, the ESS is utilised as a generator through discharge stages (on-peak). The maximum power demand from the grid (162 kw) is assigned to the ESS, domestic load of each home is 1.14 kw and wind power is kw. The power losses and voltage profile are then calculated for discharging process. The same procedure is repeated to calculate the power losses and voltage profile during the charging process. It should be noticed that during the charging process 85

5 only 162 kw are used to charge the ESSs but the extra power was allowed to be sent back to the grid. The results of using the GA and OPF on the ESSs distribution is given in Figure 10, providing the same constraint of DGs should not exceed the number of homes on the same bus. It is found that the optimum number of DGs and ESS are not similar for each busbar. This case could be looked at in details in the future as a business model. Case 1: DN has no ESSs Case 2: DN has ESSs at transformer secondary bus Case 3: DN has ESSs at the middle, bus 4 Case 4: DN has ESS at the far end bus 7 Case 5: DN has ESS distributed using GA a. During discharging: Using Maximum amount of power needed from the grid in the whole year is 162 kw in winter during discharging; Figure 13 shows the p.u Voltages of the tested systems for the five cases. Figure 14 is showing the total power losses of the tested systems under all cases. It is clear to see that voltage in cases 3, 4 and 5 is close to 1 p.u whereas it is less for cases 1 and 2. Power losses is also much less for cases 3, 4 and 5. Figure 10: Tested LV network with distributed DG and ESS using GA Impacts on voltage profile and power loss The GA and OPF has run for several trials in order to obtain an accurate results, it was noticed that almost every run gave the same results presented in Figure 10 above. The allocation of ESS at different buses with the rated values shows clear regulation of the bus voltages near its optimum value (1 p.u.) as shown in Figure 11, while it also reduces the real power losses as compared to previous publications as depicted in Figure 12. Figure13: p.u Voltages of the tested systems at different places Figure 14: Total power losses of the tested systems b. During charging: Figure 11: Voltage from using GA of 7 buses in both cases charge and discharge Figure 15: Voltages of the tested systems at different busbars Figure 12: Power losses from GA in both cases charge and discharge 3.2.Case Studies Five cases are studied to show the impacts of the ESSs and DGs on the DN performance, all cases have the optimum DGs distributed using the GA as explained in Figure 6 above; Maximum amount of unutilised wind power in the whole year is 335 kw (spring season) using power flow. Figure 15 shows the p.u Voltages of the tested systems for all five cases and Figure 16 is showing the total power losses of the same tested systems during charging, voltages in cases 1, 2, 3 and 4 are shifting from the unity in comparison to case 5, and 86

6 power losses in case 1, 2 is lower than cases 3 and 4 as shown in Figures 15 and16. Figure 16: Total power losses of the tested systems ESS has been used in different places and different size on buses. During discharging, GA has been used in case 5 to minimise power losses and improve voltage, so voltage in case 5 in both Figures 13 and 15 are close to unity, in addition power losses for the same case, has been reduced as shown in Figures 14 and 16. CONCLUSION In this paper, GA and OPF have been used to identify the optimum allocation of ESSs with DGs in distribution network. The storage energy has been used to support 100% load demand covered by renewable resources, in this case low scale wind energy has been considered. The charging and discharging scheme of the battery banks is optimized to minimise the number of the wind turbines used in the distribution network and to minimise the capacity of the total battery banks. The results showed that the target of minimising the capacity of ESSs and WT number is achieved in addition to maintain a steady voltage profile across all busbars and minimisation of the power losses. REFERENCES [1] A. Alhamali, E. Mohamed Farrag, B. Geraint, and D. Hepburn. "Review of Energy Storage Systems in electric grid and their potential in distribution networks." In Power Systems Conference (MEPCON), 2016 Eighteenth International Middle East, pp IEEE, [2] A. Chatzivasileiadi, E. Ampatzi, and I. Knight. "Characteristics of electrical energy storage technologies and their applications in buildings." Renewable and Sustainable Energy Reviews 25 (2013): [3] A.K Srivastava, A.A. Kumar, and N.N. Schulz. "Impact of distributed generations with energy storage devices on the electric grid." Systems Journal, IEEE 6, no. 1 (2012): [4] The APS Panel on Public Affairs Committee on Energy and Environment, Challenges of Electricity Storage Technologies, [5] Chatzivasileiadi, Katerina. "Electrical energy storage technologies and the built environment." In International Refdbnewable Energy Storage Conference. [6] [6] P.M. Van de Ven, N. Hegde, L. Massoulié, and T. Salonidis. "Optimal control of end-user energy storage." Smart Grid, IEEE Transactions on 4, no. 2 (2013): [7] M. Zidar, P.S. Georgilakis, N.D. Hatziargyriou, T.Capuder, and D. Škrlec. "Review of energy storage allocation in power distribution networks: applications, methods and future research." IET Generation, Transmission & Distribution (2015). [8] J. Xiao, Z. Zhang, L. Bai, and H. Liang. "Determination of the optimal installation site and capacity of battery energy storage system in distribution network integrated with distributed generation." IET Generation, Transmission & Distribution (2015). [9] N.A. Ashtiani, M. Gholami, and G. B. Gharehpetian. "Optimal allocation of energy storage systems in connected microgrid to minimize the energy cost." In Electrical Power Distribution Networks (EPDC), th Conference on, pp IEEE, [10] A. S. Awad, T. H. EL-Fouly, and M. M. Salama, Optimal ESS allocation for load management application, IEEE Trans. Power Syst., vol. 30, no. 1, pp , Jan [11] M. F. Shaaban, Y. M. Atwa, and E. F. El-Saadany, DG allocation for benefit maximization in distribution networks, IEEE Trans. Power Syst., vol. 28, no. 2, pp , May [12] G. Carpinelli, F. Mottola, D. Proto, and A. Russo. "Optimal allocation of dispersed generators, capacitors and distributed energy storage systems in distribution networks." In Modern Electric Power Systems (MEPS), 2010 Proceedings of the International Symposium, pp IEEE, [13] G. Carpinelli, S. Mocci, F. Mottola, F. Pilo, D. Proto. Optimal integration of distributed energy storage devices in smart grids, IEEE Trans. Smart Grid, 2013, 4, (2), pp [14] G. Celli, S. Mocci, F, Pilo, M. Loddo. Optimal integration of energy storage in distribution networks IEEE Bucharest PowerTech, [15] M. Zidar, P.S. Georgilakis, N.D. Hatziargyriou, T. Capuder, and D. Škrlec. "Review of energy storage allocation in power distribution networks: applications, methods and future research." IET Generation, Transmission & Distribution, [16] D. Yao, S. Choi, K.J. Tseng, T. Lie. A statistical approach to the design of a dispatchable wind power-battery energy storage system, IEEE Trans. Energy Convers., 2009, 24, (4), pp [17] Y.M. Atwa, and E. F. El-Saadany. "Optimal allocation of ESS in distribution systems with a high penetration of wind energy." Power Systems, IEEE Transactions on 25, no. 4 (2010): [18] M. Motalleb, E. Reihani, and R. Ghorbani. "Optimal placement and sizing of the storage supporting transmission and distribution networks."renewable Energy 94 (2016): [19] A. Alhamali, M. E. Farrag, B. Geraint, and D. Hepburn. " Determination of Optimal Site and Capacity of DG Systems in Distribution Network based on Genetic Algorithm." In Power engineering conference Conference (UPEC), st International Universities Power Engineering Conference. [20] Met Office Library & Archive (2016). Hourly wind data for Glasgow. Available: [21] P. Suwanapingkarl. "Power quality analysis of future power networks." PhD diss., Northumbria University, [22] M. Ghofrani, A. Arabali, M. Etezadi-Amoli and M.S. Fadali. "A framework for optimal placement of energy storage units within a power system with high wind penetration." IEEE Transactions on Sustainable Energy 4, no. 2 (2013):