Abstract: Aiming at the problems of DC bus voltage and load fluctuation and unstable transmission power of dual active bridge(DAB) converter in DC microgrid system, a hybrid optimal control strategy based on genetic algorithm’s self-resilient control and gradient descent algorithm to optimize the return power is proposed. Firstly, the topology and power characteristics of the DAB converter under extended phase-shift modulation are analyzed, and the gradient descent algorithm is introduced to find the optimal inward shift phase ratio iteratively by comparing the backflow power as a loss function. Secondly, based on the small-signal modeling of the DAB converter, a linear self-immunity controller is designed to observation and estimation the output voltage and internal and external disturbances of the system through the observation of the expanded state observer. At the same time, considering the uncertainty of parameter tuning of the self-immunity controller in the complex environment, a genetic algorithm is introduced to self-tune the parameters of the self-immunity controller. Finally, an experimental platform with TMS320F28335 as the controller is built to analyze and compare the hybrid optimal control strategy under extended phase shift modulation(EPS-HOCS), PI control strategy under extended phase shift modulation(EPS-PI) and adaptive gradient descent algorithm under extended phase shift modulation(EPS-AGDA), which verifies the superiority of the proposed strategy in terms of return power and dynamic performance.