受脑启发的协作式自动发电控制与大规模电动汽车集成

Distributed energy sources, loads and storage equipment have intermittent and highly random characteristics, which can cause significant frequency fluctuations when they are integrated into a distributed power grid. The current multi-agent cooperative neural networks based algorithms in the distributed power grid would suffer from catastrophic forgetting issues, which might be difficult in achieving optimal control under strong random disturbances. Hence, this paper proposes a proximal brain-inspired policy optimization (PBPO) algorithm with an orthogonal weight modification method in the weight update of the policy networks. Thus the policy network can have the brain-inspired contextual awareness capability. It can obtain faster convergence to the optimal solution of the multi-area cooperative control, mitigating heavy frequency fluctuations caused by serious random disturbances in the grid. The effectiveness of the proposed algorithm is validated via the simulation experiments on the two load frequency control models of the energy integration of large-scale electric vehicles in the grid. The proposed PBPO algorithm outperforms various reinforcement learning algorithms with faster convergence, higher frequency stability and better control performance.