Deep Reinforcement Learning Based Input Voltage Sharing Method for Input-Series Output-Parallel Dual Active Bridge Converter in DC Microgrids

The input-series output-parallel connected dual active bridge (ISOP-DAB) converter is an attractive solution to connect medium-voltage dc (MVdc) and low-voltage dc (LVdc) grids. This paper proposes an input voltage sharing (IVS) control algorithm for a multi-agent (MA) ISOP-DAB converter based on the deep reinforcement learning (DRL) method. Compared with other methods, the proposed control algorithm can regulate the output voltage and ensure the IVS of the ISOPDAB converter adaptively in real-time. Real-time simulations in OP5600 validate that the proposed algorithm has good dynamic performance.