A Fast Converged Voltage Control Method based on Deep Reinforcement Learning

Keeping the voltage level in desired range is of great importance to power grid’s stability. Thus, the voltage control is one of the most important task in daily power grid’s operation. However, with the integration of distributed energy resources, demand response mechanism, and new devices, ensuring the stability of power grid becomes a challenging task. The traditional secure and economic operation, which is defined off-line, should be re-determined. In this article we propose a deep reinforcement learning based fast converged autonomous voltage control method, along with well-defined training paradigm and fine-tuned hyper parameter. A IEEE 14-bus case is used for testing the applicability and performance of the method. The result shows that our method has enough capability of controlling the voltage level in power grid with high convergence speed.