Artificial intelligence based approach to improve the frequency control in hybrid power system. (December 2020)

• Record Type: Journal Article
• Authors: Wang, Hao; Zhang, Guozhou; Hu, Weihao; Cao, Di; Li, Jian; Xu, Shuwen; Xu, Dechao; Chen, Zhe
• Is Part Of: Energy reports. Volume 6(2020)Supplement 8
• Page Start: 174
• Page End: 181
• Publication Date: 2020-12
• Subjects: Frequency support -- MTDC -- Droop control -- Deep reinforcement learning; Power resources -- Periodicals; Energy industries -- Periodicals; Power resources; Periodicals; Electronic journals; 621.04205
• DOI: 10.1016/j.egyr.2020.11.097
• Languages: English
• ISSNs: 2352-4847

Abstract

Abstract: Frequency control over networks is done using the frequency droop control technique which has the simplicity advantage although it allows that, in certain situations, frequency control is not very efficient. Artificial intelligence techniques have been increasingly used, so it is justified to explore their viability in electrical networks. The present work analyzes the use of Artificial Intelligence in networks to improve the frequency droop control. In order to realize this, a deep reinforcement learning (DRL)-based agent is proposed to tune the controller parameters for voltage source converter (VSC) in this paper. The DRL-based agent is trained by numerous hybrid grid operation conditions to lean the optimal control policy, which make it achieve a good adaptability to variety of operation conditions. For the purpose of demonstrating this method, a time-domain simulation model of hybrid power system is built with MATLAB/Simulink to act as test system. The simulation results verify the effectiveness of the proposed method.