Deep reinforcement learning-driven decision-making paradigm for electricity-carbon-hydrogen collaboration

To achieve synergistic optimization of low-carbon energy systems, electricity-carbon-hydrogen synergy has become one of the critical pathways. However, its high dimensionality, nonlinearity, and strong uncertainties limit traditional optimization methods. Deep reinforcement learning (DRL), with its ability to learn from data, adapt to dynamic environments, and support multi-objective decision-making, offers a promising solution. This paper reviews the mechanisms of electricity-carbon-hydrogen synergy and the necessity of applying DRL, summarizing recent progress in electricity markets, carbon markets, electricity-carbon synergy, electricity-hydrogen synergy, and their integration. The results show that DRL holds significant potential for enhancing renewable energy integration, optimizing carbon trading, and coordinating multi-energy flows, though challenges remain in model complexity, interpretability, safety, and multi-objective trade-offs. Future research should focus on integrating DRL with large language models, improving robustness, safety, and interpretability, and enabling cross-scale coordination to facilitate practical deployment.