Multiagent deep meta reinforcement learning for sea computing-based energy management of interconnected grids considering renewable energy sources in sustainable cities

In a sustainable city with a large amount of renewable energy, the difficulty inherent in the coordination of the load frequency control strategies of grid units and area units lead to severe frequency fluctuations. Conventional load frequency control has difficulty overcoming the above problems due to its lack of adaptive control ability and robustness. To overcome this problem, this paper proposes a sea computing-based grid-area coordinated load frequency control (SCGAC-LFC) method, which equates each grid unit and area unit in each area as independent agents. Instead of different units relying on different strategies, all units collectively obtain the LFC policy that suits the market requirements. In its online application, no communication is needed since each unit can arrive at its own decision. In addition, this paper proposes a curriculum multiagent deep meta-actor-critic (CMA-DMAC) algorithm, which introduces meta-reinforcement learning and curriculum learning to guide the agent training to improve the robustness and quality of the obtained SCGAC-LFC strategies. Using a simulation of the four-area LFC model for the China Southern Grid (CSG), our proposed method carries significantly lower frequency error and regulation mileage payment.

Automated summary

This study proposes a sea computing-based grid-area coordinated load frequency control method which effectively reduces frequency fluctuations and improves quality and robustness. By introducing meta-reinforcement learning and curriculum learning, the method guides agent training to obtain LFC strategies that suit market requirements.