Improving DFIG Performance under Fault Scenarios Through Evolutionary Reinforcement Learning Based Control

Wei Gao,Rui Fan,Renke Huang,Qiuhua Huang,Yan Du,Wei Qiao,Shaobu Wang,David Wenzhong Gao
DOI: https://doi.org/10.1049/gtd2.12563
2022-01-01
Abstract:The doubly fed induction generator (DFIG) usually experiences high rotor current and DC capacitor link voltage spikes during system fault events. In this paper, a novel data-driven approach is proposed to enhance DFIG performance under fault scenarios. An advanced reinforcement learning algorithm called guided surrogate-gradient-based evolution strategy (GSES) is used to control the DFIG power and capacitor DC-link voltage by adjusting the optimal reference signals. This controller is able to prevent the DFIG rotor from over-current risk and maintain grid-connected operation. The proposed GSES-based control algorithm was evaluated through simulations on a 3.6-MW DFIG in the PSCAD/EMTDC software. Results have validated the effectiveness of the proposed GSES-based control algorithm in improving DFIG performance under various fault scenarios.
What problem does this paper attempt to address?