Near-Optimal Energy Management Strategy for a Grid-Forming PV and Hybrid Energy Storage System

Integration of Li-ion batteries and supercapacitors (SCs) into PV plants enables a hybrid PV system with more grid functions like power filtering and frequency regulation. Above that, an energy management system (EMS) plays a key role in achieving grid functions and economic performance. However, previous efforts focused on advanced forecast methods without considering real-time EMS. This paper thus aims to develop a practical real-time EMS with near-optimal performance for the degradation of the hybrid energy storage system (HESS). Firstly, a variational mode decomposition (VMD) method is combined with a long short-term memory (LSTM) network to decompose and learn feature parameters of typical historical weather data, improving forecast accuracy and shifting the operation mode periodically. Then, the mixed integer linear programming approach is utilized to find out the optimal control mode in different operation scenarios, and three-segment rules are extracted from the optimization results. Finally, the deep learning-based real-time EMS is developed. Numeric simulations validate that the proposed EMS can achieve near-optimal performance with a low computation burden. Besides, the proposed strategy can reduce the degradation cost by up to 80% compared with competitive rule-based strategies.