Deep-reinforcement-learning-based Primary Frequency Regulation Control for Hybrid Renewable Power System

Abstract

With the increasing penetration of renewable energy sources, the frequency control strategy of power system completely depending on traditional power supply becomes unsustainable. This objective situation requires renewable energy sources take part in the frequency regulation. This paper proposes a new primary frequency regulation (PFR) control based on DDQN algorithm, which makes full use of renewable energy ’ s fast response ability. Two primary frequency regulation control on DDQN algorithm for hybrid renewable power system is presented in this paper : one uses DDQN algorithm to optimize droop gain; the other uses DDQN algorithm to directly optimize the power output of each power generation. The presented method is validated with a actual hybrid renewable power system in Liaoning province. The experiment results demonstrates that power output control based on DDQN algorithm obviously improves the frequency characteristics of the grid. Compared with traditional frequency regulation, proposed control strategy can effectively improve lowest frequency point in primary frequency regulation process. The experiment shows DDQN algorithm has great potential in primary frequency regulation of hybrid renewable power system.