面向电-氢耦合系统的双环运行优化框架：基于模型与无模型联合方法

Recently, electric–hydrogen coupled systems (EHCS) have gained wide attention. However, the spatiotemporal nonlinear constraints arising from the coupling and interaction of electricity and hydrogen energy flows, together with the persistent disturbances caused by multiple uncertainties from sources, loads, and devices, pose significant challenges to their operational optimization. Accordingly, we propose a dual–loop optimization framework with a joint solving loop (inner loop) and a knowledge correction loop (outer loop). The inner loop employs ADMM to structurally decompose the system optimization into electricity and hydrogen subproblems, mitigating solution–space complexity. It then matches model–based optimization (with its sample–efficiency advantage) to the electricity chain’s quasi–static behavior at the decision timescale, and model–free learning (with its long–horizon decision robustness advantage) to the hydrogen chain’s dynamics and uncertainties, enabling efficient coordinated solving across both chains. The outer loop derives implicit rewards from operational features and expert knowledge to dynamically correct the explicit cost–based reward, providing prior guidance that improves the agent’s learning efficiency. The two loops operate in coordination, forming a customized dual–loop optimization paradigm tailored for coordinated operation of the electricity and hydrogen chains, with the goal of balancing theoretical optimality and practical execution robustness. Simulation results demonstrate the effectiveness of the proposed method.