Abstract
To eliminate voltage violations and improve the security and economy of wind farms, model-based volt/var optimization has been widely but heavily relies on accurate wind farm models. However, numerous wind farms lack of well maintained models because of rapid construction, limited operation staff and changing environment. It motivates us to develop a data-driven wind farm volt/var control based on deep reinforcement learning (DRL). In order to improve the stability, efficiency and optimality of existing DRL-based methods, entropy-regulation technique is utilized. Numerical simulations not only demonstrate the effectiveness of such data-driven volt/var control method, but also show the superiority of our proposed method.
Haotian Liu
PhD of Electrical Engineering
My research interests include reinforcement learning, contextual bandit, power system and model-free control.