Artificial Intelligence Data Driven Control for DC Solid State Transformer in DC Microgrid

The DC solid-state transformer (DCSST) serves as a key component for connecting power supplies, loads, and other elements in a DC microgrid system. With the integration of new energy sources and random loads, the system has become increasingly complex, making accurate modeling challenging due to high modeling costs. The randomness of the loads introduces various disturbances, compromising system stability. To address these challenges, this paper proposes a data-driven model-free control (DMFC) approach based on deep reinforcement learning (DRL) and an improved Kalman filter (IKF) for the DCSST. First, a data-driven model of the DCSST is established. Next, the IKF is designed, which is tailored for DMFC to correct measurement errors, filter noise, and estimate uncertain disturbances. Subsequently, the DMFC is introduced, which ensures stability under large signal disturbances without relying on an accurate system model. Finally, a DRL sub-controller is integrated into the DMFC to intelligently adjust the control signal and enhance the system’s operational adaptability. Experimental results demonstrate that the proposed control method can adaptively ensure the stability of the system under large signal disturbances without requiring an accurate model, while also exhibiting good dynamic and steady-state performance.