An Auxiliary Strategy for Real-time Optimal Dispatch of Microgrids Based on XGBoost Machine Learning Algorithm

Abstract:
Real-time optimal dispatch is an important part of microgrid operation, which is of great significance to the safety and stability of system. With the access of large-scale uncontrollable renewable energy resources, microgrid is gradually evolving into a new system with openness, complexity and uncertainty. The traditional model-driven optimal dispatch strategy has been difficult to meet the actual needs. Machine learning algorithm provides a new solution for system scheduling. This paper proposes a novel auxiliary strategy for real-time optimal dispatch of microgrids based on XGBoost algorithm. Specifically, the historical operating data of microgrid is used to establish a XGBoost-based dispatch model, which can directly construct the relationship between the operating status and the dispatch decision so as to provide the auxiliary strategy quickly. The grid-connected microgrid located in Guigang, Guangzhou is taken as a case for verification, and the optimal dispatch results of DNN algorithm and LSTM algorithm are compared and analyzed. Simulation results prove the effectiveness of the proposed model and algorithm.