Optimal FNN-Based Energy Management System With High Real-Time Performance and Good Interpretability for Battery in Grid-Connected Microgrid

作     者：Liu, Bin Wang, Dan Huang, Jiawei Mao, Chengxiong 

作者机构：Huazhong Univ Sci & Technol State Key Lab Adv Electromagnet Technol Wuhan 430074 Peoples R China 

出 版 物：《IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS》 (IEEE Trans Ind Electron)

年 卷 期：2025年第72卷第8期

页      面：8142-8153页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0804[工学-仪器科学与技术] 0811[工学-控制科学与工程] 

主　　题：Energy management Microgrids Real-time systems Fuzzy neural networks Artificial neural networks Fuzzy control Renewable energy sources Costs Optimization models Wind speed Fuzzy neural network (FNN) high real-time performance interpretability microgrid optimal energy management 

摘      要：In this article, a novel energy management system (EMS) is presented for a grid-connected microgrid using the fuzzy neural network (FNN), aiming to minimize the mismatch between renewable power generation and load power demand in the microgrid by controlling the battery charge/discharge power in real time, so as to effectively promote the local consumption of renewable energy. An on-line FNN controller is used to rapidly generate energy management instructions in response to the random variations of the net load power in real time, where the parameters are updated through off-line training periodically. The simulation results show that: 1) the presented FNN-based EMS can get better optimization results compared to each benchmark EMS, achieving an average decrease of 18.0217% on the optimization function value in all tested cases when regarding the best performing one among benchmark EMSs as the comparison object;2) the presented FNN-based EMS has high real-time performance on level of seconds;and 3) the presented FNN-based EMS has good interpretability that all the used parameters in the FNN have interpretable meanings. The experimental results on the testbed match well with the corresponding simulation results, demonstrating the effectiveness and practicability of the presented FNN-based EMS for practical applications.