Power balance control of RES integrated power system by deep reinforcement learning with optimized utilization rate of renewable energy

Abstract

A power balance control method is proposed for renewable energy source (RES) integrated power systems based on deep reinforcement learning (DRL), with the reasonable utilization rate of renewable energy optimized. This method considers the dispatching problems of a high-proportion renewable energy grid from a new perspective. It is proposed that the dispatching of a high-proportion renewable energy grid must consider the reasonable utilization rate of renewable energy and conduct reasonable abandonment of wind and light. And in the offline training of DRL scheduling, the reasonable utilization rate is used as the element of the state vector to train the final power grid DRL control strategy. The control strategy has verified its effectiveness in the IEEE 14-bus system with the supporting datasets.