 This paper proposes a novel approach for optimizing the operation of a multi-microgrid system, consisting of multiple renewable energy sources. It utilizes a multi-agent centralized training, distributed execution framework, to enable energy trading between the microgrids. Furthermore, it employs automated machine learning to optimize the hyperparameters of a deep reinforcement learning algorithm to improve its performance. The results show that the proposed method can achieve power complementarity between the microgrids and reduce their operating costs. Additionally, it outperforms other state-of-the-art reinforcement learning algorithms with better economy and higher calculation efficiency. This article was authored by Jian Kai-Gao, Yang Li, Bin Wang, and others.