Research on load frequency control of multi‐microgrids in an isolated system based on the multi‐agent soft actor‐critic algorithm

Load variation, distributed power output uncertainty and multi-microgrids network complexity have brought great difficulties to the frequency stability of whole microgrid. To address this problem, paper uses a multi-agent deep reinforcement learning(DRL) algorithm design controllers control multi-microgrids. Firstly, load (LFC) model for was built. Secondly, based on centralized training decentralized execution (CTDE) learning (RL) framework, soft actor-critic (MASAC) designed applied model. The state space action were established according deviation every sub-microgrid each source. reward function then deviation. appropriate neural parameters selected generate interconnected microgrid through multiple pre-learning. Finally, simulation study shows that MASAC controller proposed in can quickly maintain when system is disturbed. Sensitivity analysis effectively cope with parameters.