In island mode, microgrids need to operate independently from traditional power systems, efficiently coordinating internal ener?gy to ensure the continuity and efficiency of energy supply. The twin delayed deep deterministic policy gradient algorithm significantly im?proves the processing efficiency and accuracy of complex continuous control tasks through policy delay updates and the introduction of du?al Q networks. Based on this way, an energy optimization allocation strategy is designed for microgrids embedded with fuel cells based onthe TD3 algorithm, to improve the stable power supply capacity and quality of the microgrid system, reduce energy consumption and opera?tion costs, and enhance the system’s economy and reliability. Through comprehensive analysis, the comprehensive performance of the de?signed energy optimization allocation strategy in different scenarios is comprehensively evaluated. The results show that by optimizing thecharging and discharging modes and ratios of fuel cell systems, the energy optimization allocation strategy designed based on TD3 algo?rithm performs better than traditional algorithms in improving energy allocation efficiency, shortening response time, and reducing operat?ing costs. The research results have verified the efficient adaptability of TD3 algorithm in dealing with fluctuations in renewable energygeneration power output and changes in load demand, and it has wide applicability in practical energy management scenarios.