Reinforcement-learning-based voltage-frequency control of microgrids under non-ideal situations

Due to the currently high penetration rate and low inertia of renewable energy, microgrid (MG) stability requirements are gradually increasing. Non-ideal situations (such as line failure and network attack) will especially seriously affect the safe operation of the system. This paper proposes a secondary control problem of MGs under variable working situations, designs a voltage-frequency control strategy utilizing the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, and provides a detailed description of the design process. Through simulation experiments, the TD3 algorithm is pre-learned offline and then tested online in two situations. Compared with other strategies, the effectiveness and adaptability of the proposed algorithm are verified.