In this paper, we propose a data-driven load frequency control(LFC) method for a wind-fire multi-energy complementary power generation system. The method converts the LFC problem into a maximization reward function problem by constructing a reward function that includes control performance criteria (CPS) and dynamic performance indicators. It introduces a deep deterministic policy gradient(DDPG) algorithm to solve the problem. Then, the optimal adaptive coordinated frequency control strategy under the actual wind turbine output is obtained through pre-learning and online application. Finally, the analysis is conducted to evaluate the mid-to long-term control performance. The effectiveness and feasibility of the proposed method in improving the performance of LFC are verified by conducting a simulation with continuous step disturbance perturbations. The simulation results show that the proposed algorithm can effectively suppress fluctuations when the system is perturbed, and reduce the regulation time required to complete LFC.

Load frequency control,multi-energy complementary power generation system,maximization reward function problem,deep deterministic policy gradient,CPS index