Greedy energy management strategy and sizing method for a stand-alone microgrid with hydrogen storage

Abstract

This paper presents a greedy energy management strategy based on model predictive control (MPC) for a stand-alone microgrid powered by photovoltaic (PV) arrays and equipped with batteries and a power-to-hydrogen-to-power (P2H2P) system. The proposed strategy consists of a day-ahead plan and an intra-day dispatch method. In the planning stage, the sequence of plan is to determine the power of each storage device for a certain period, which is initially generated under the principle that PV arrays have the highest priority, followed by the batteries, and finally the P2H2P system, using short-term forecast data of both load and solar irradiance. The initial plan can be optimized with objectives of harvesting more PV generation in storage and minimizing unmet load through rescheduling P2H2P system and batteries. Three parameters including reserved capacity of batteries, predischarge coefficient of fuel cell (FC), and greedy coefficient of electrolyzer (EL), are introduced during plan optimization process to enhance the robustness against forecast errors. In the dispatching stage, the energy dispatch is subject to the scheduled plan and the operational constraints. To demonstrate the capabilities of the proposed strategy, a case study is performed for a hotel with a mean power consumption of 1567 kWh/day based on the system configuration optimized by HOMER software in comparison with the load following (LF) strategy and the global optimum solution solved by mixed integer linear programing (MILP). The simulation results show that the annual unmet load using the proposed strategy is reduced from 13,434 kWh to 2370 kWh, which is 528 kWh lower than the optimum solution. Meanwhile, the cost of energy (COE) of the proposed strategy decreases by US$ 0.08/kWh compared to the LF strategy and is equal to the optimum solution. Finally, the performance of configuration optimization employing genetic algorithm (GA) under different energy management strategies is investigated with the objective function of minimizing the net present cost (NPC). Furthermore, the robustness of the proposed strategy is studied. The results show that the proposed strategy gives an NPC and COE of US$ 2.4 million (Mn) and US$ 0.43/kWh, which are 23.4% and 9.7% lower than those of systems utilizing the SoC-based strategy and the LF strategy, respectively. The results also demonstrate that the strategy is robust against forecast errors, especially for overestimated forecast models.