Аннотация
This paper presents a comprehensive energy management framework for a sustainable waste valorization system (SWVS) integrating electricity, heat, cooling, natural gas, hydrogen, and waste‐to‐energy infrastructures within a unified multi‐energy system (MES). The proposed configuration incorporates sustainable resources (wind turbines [WTs] and photovoltaic [PV] units), alongside combined heat and power (CHP), waste‐to‐energy plants (WTPs), multi‐vector energy storage technologies, carbon capture, utilization, and storage (CCUS), and methanation reactors to enable circular carbon utilization. A detailed nonlinear mathematical model is developed to capture realistic operational characteristics, including coupled electricity‐heat production, multi‐carrier storage dynamics, pollutant emissions (CO 2 , SO 2 , and NO 2 ), and energy balance constraints across all carriers. To address the high dimensionality, strong nonlinear coupling, and uncertainty inherent in SWVS operation, an advanced energy management strategy based on DeepSeek with Group Relative Policy Optimization (GRPO) is developed. Unlike conventional actor‐critic approaches, the DeepSeek‐GRPO method eliminates value‐function approximation and directly optimizes policies through multi‐action sampling and reward‐based advantage estimation, leading to enhanced training stability, scalability, and constraint adherence. The proposed framework simultaneously minimizes total operational cost and environmental emissions while satisfying technical and operational constraints. Comprehensive analysis demonstrates that the proposed DeepSeek‐GRPO approach provides a high level of scheduling operation for integrated sustainable waste valorization and MES operation.