Computational intelligence techniques for energy storage management

Ph. D. ThesisThe proliferation of stochastic renewable energy sources (RES) such as photovoltaic and wind power in the power system has made the balancing of generation and demand challenging for the grid operators. This is further compounded with the liberalization of electricity market and the introduction of real-time electricity pricing (RTP) to reflect the dynamics in generation and demand. Energy storage sources (ESS) are widely seen as one of the keys enabling technology to mitigate this problem. Since ESS is a costly and energy-limited resource, it is economical to provide multiple services using a single ESS. This thesis aims to investigate the operation of a single ESS in a grid-connected microgrid with RES under RTP to provide multiple services. First, artificial neural network is proposed for day-ahead forecasting of the RES, demand and RTP. After the day-ahead forecast is obtained, the day-ahead schedule of energy storage is formulated into a mixed-integer linear programming and implemented in AMPL and solved using CPLEX. This method considers the impact of forecasting errors in the day-ahead scheduling. Empirical evidence shows that the proposed nearoptimal day-ahead scheduling of ESS can achieve a lower operating cost and peak demand. Second, a fuzzy logic-based energy management system (FEMS) for a grid-connected microgrid with RES and ESS is proposed. The objectives of the FEMS are energy arbitrage and peak shaving for the microgrid. These objectives are achieved by controlling the charge and discharge rate of the ESS based on the state-of-charge (SoC) of ESS, the power difference between RES and demand, and RTP. Instead of using a forecasting-based approach, the proposed FEMS is designed with the historical data of the microgrid. It determines the charge and discharge rate of the ESS in a rolling horizon. A comparison with other controllers with the same objectives shows that the proposed controller can operate at a lower cost and reduce the peak demand of the microgrid. Finally, the effectiveness of the FEMS greatly depends on the membership functions. The fuzzy membership functions of the FEMS are optimized offline using a Pareto based multi-objective evolutionary algorithm, nondominated sorting genetic algorithm- II (NSGA-II). The best compromise solution is selected as the final solution and implemented in the fuzzy logic controller. A comparison was made against other control strategies with similar objectives are carried out at a simulation level. Empirical evidence shows that the proposed methodology can find more solutions on the Pareto front in a single run. The proposed FEMS is experimentally validated on a real microgrid in the energy storage test bed at Newcastle University, UK. Furthermore, reserve service is added on top of energy arbitrage and peak shaving to the energy management system (EMS). As such multi-service of a single ESS which benefit the grid operator and consumer is achieved