Joint Energy and Workload Scheduling for Fog-Assisted Multimicrogrid Systems: A Deep Reinforcement Learning Approach
Tingjun Zhang, Dong Yue, Liang Yu, Chunxia Dou, Xiangpeng Xie
Abstract
Due to the fluctuation of renewable energy, the uncertainty of electrical loads, and the complexity of networked microgrids, it is challenging to dispatch multiple resources to minimize the operating cost of multi-microgrid system (MMGS). In this article, a fog-assisted operating cost minimization problem of MMGS is investigated with the consideration of source-grid-load-storage-computing coordination. Since there are strong couplings among multiple resources, it is difficult to solve the problem directly. Therefore, the above problem is reformulated as a Markov game. Then, a novel energy management algorithm is proposed to solve Markov game based on multiagent deep reinforcement learning. It is worth mentioning that the proposed energy management algorithm can support local real-time decisions for each microgrid without knowing any prior knowledge of uncertain parameters and private information of other microgrids. Simulation results indicate that the proposed algorithm can reduce the long-term cost of each microgrid by 0.09 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> –8.02 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> compared with baselines.