Volume 32: A Sustainable, Clean and Carbon-free Energy Future

Deep reinforcement learning based adaptive energy management for islanded microgrids considering multi-objective optimization Jiawen Li, Tao Zhou

Abstract

In order to reduce the frequency deviation and unit generation cost of an isolated microgrid, an adaptive load frequency control (ALFC) method is proposed in this paper. The method employs an adaptive proportional-integral-derivative (PID) controller to achieve adaptive control by adaptively adjusting the parameters of the controller to output the regulation command. In addition, to achieve adaptive regulation of the control parameters, a deep actor-critic (DAC) algorithm is proposed in this paper, which introduces multiple critics and Gaussian noise exploration techniques to enhance the quality of the ALFC strategy. The performance of the proposed method is tested in the Zhuzhou isolated microgrid of the China Southern Grid(CSG), which can effectively reduce frequency deviation and generation cost.

Keywords energy management, load frequency control, deep reinforcement learning, frequency deviation, generation cost