Prior Knowledge Incorporated Large-Scale Multiagent Deep Reinforcement Learning for Load Frequency Control of Isolated Microgrid Considering Multi-Structure Coordination
Jiawen Li, Tao Zhou
Abstract
In load frequency control (LFC) of island microgrids, the objectives of the controller and power distributor are inconsistent, which increases the frequency deviation and total generation cost. To solve this problem, a data-driven multi-input multioutput LFC (MIMO-LFC) method is proposed. Furthermore, an innate-oriented large-scale multiagent double delayed deep deterministic policy gradient algorithm is proposed for this method. The design of the algorithm is based on a human cognitive mechanism, whereby <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a priori</i> knowledge is delivered to the agent before prelearning to guide the learning of the agent more effectively, thus improving the robustness of MIMO-LFC. This method integrates the controller and distributor into one agent and solves the cooperative control problem arising between the controller and power distributor by outputting the command of each unit directly. An experiment on the Zhuzhou Island microgrid verifies the effectiveness of the proposed method.