Rejection Depth Differential Dynamic Programming for Real-Time Integrated Control for Micro-Grids

With the continuous addition of various heterogeneous energy sources and power users in the micro-grid, it causes different degrees of impacts on the operational stability of the grid and highlights the inadequacy of traditional control strategies in frequency control and economic scheduling of micro-grids, therefore, it is very necessary to study the economic dispatch and frequency control of multiple distributed new energy sources after they are connected to the micro-grid. Although the traditional deep learning controller can provide the output, it will lead to the deviation of the micro-grid system when the prediction probability is not high. Based on the rejection operation, this paper presents a framework for real-time generation scheduling and control of micro-grids. When the prediction probability is not high enough, the denial of knowledge deep differential dynamic programming can provide the action of analyzing the control algorithm; when the prediction probability is lower than the set probability cutoff, the exportation of the traditional proportional-integral-derivative method is selected as the output of the controller. The numeric simulation results show that the proposed algorithm replaces the traditional micro-grid framework, and establishes the rejection deep differential dynamic planning method for micro-grids.