부산대학교 도서관

This paper develops a novel power system emergency control scheme to improve short-term voltage stability (STVS) using deep reinforcement learning (DRL). The proximal policy optimization (PPO) algorithm is developed to adaptively update the load shedding strategy and the deep neural network parameters through continuously interacting with the environment. In order to assess the effects of the load shedding action, a reward function is designed by using a transient voltage severity index. Simulations are conducted on a modified New England 39-bus system, and the results illustrate that the DRL algorithm can obtain a better strategy in a variety of STVS scenarios.