Evolutionary‐Game‐Informed Adaptive MAPPO for Risk Prediction and Control in Coupled Electricity‐Carbon‐Green Certificate Markets

The coordinated operation of electricity, carbon, and green certificate markets creates a coupled decision environment with strong nonlinearity, bounded‐rational agent interactions, and cross‐market risk propagation. Existing studies either rely on static game analysis or apply generic reinforcement learning methods without explicitly embedding multi‐market risk transmission into the decision process. To address this gap, this paper develops an evolutionary‐game‐informed Adaptive Multi‐Agent Proximal Policy Optimization (A‐MAPPO) framework for risk prediction and control in coupled electricity‐carbon‐green certificate markets. First, a coupled market risk model is established to characterize electricity price risk, carbon trading tail risk, and green certificate price‐demand risk, as well as their cross‐market transmission under bounded rationality. Second, these risk states are embedded into the A‐MAPPO framework through a risk‐sensitive reward function, adaptive learning and exploration schedules, and dynamic reward weighting, enabling coordinated strategy evolution and risk‐aware policy updates. Third, Tianjin market operation data from 2022 to 2024 are used as a real‐world validation case. The results show that the proposed framework can reproduce observed market restructuring and green‐power growth while improving price stability, reducing transaction frictions, and enhancing the controllability of coupled‐market risks. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.