Carbon Assessment and Carbon Reduction Potential Analysis of Rural Integrated Energy System Based on Reinforcement Learning

Recent years have seen Reinforcement Learning (RL) excel in real-time decision-making and coordinated optimization of complex energy systems. Scholars have applied RL to carbon emission optimization in urban microgrids and industrial parks, but a gap exists in rural integrated energy systems (RIES) regarding carbon assessment and reduction potential analysisrural energy data is coarse and variable, and rural social-technical coupling is complex, demanding RL for dynamic trade-offs. This study introduces an RL framework to build a "data-driven-model collaboration-strategy evolution" system, uses heuristic algorithms for RIES carbon accounting, scenario simulation and reduction evaluation, compares with algorithms like greedy to find the optimal one, and supports precise carbon reduction and rural revitalization. Results show that the proposed heuristic algorithm reduces costs by 32.3% and carbon emissions by 32.1% compared to the pure diesel algorithm, and balances economic efficiency and low-carbon benefits more effectively than the greedy algorithm.