Abstract: To address the problems of difficult multi-objective coordination, low efficiency of evolutionary algorithms, and insufficient exploration capability in the operation optimization of electricity-hydrogen-thermal integrated energy systems, a bi-objective operation optimization method based on the Double Deep Q-Network (DDQN) assisted Third-Generation Non-dominated Sorting Genetic Algorithm (NSGA-III) is proposed. The adaptive selection of genetic operators is modeled as a Markov decision process, and an agent is trained within the DDQN framework to adaptively select the optimal operator according to the population evolution state, which is then combined with the reference point selection mechanism of NSGA-III to obtain a superior Pareto front. Simulation results indicate that the proposed method is superior to comparison schemes in terms of convergence and solution set quality, and the system operating costs are reduced by 1.1%–6.3% at the same carbon emission level. By enhancing the evolutionary efficiency and global exploration capability of the algorithm, a superior Pareto front strategy is obtained, and an effective trade-off between the economic efficiency and low-carbon property of the energy system is achieved.