System construction and theoretical analysis of green power life-cycle accounting based on large language models

Abstract

Green power from renewable resources is essential for low-carbon development, but its life-cycle management is constrained by fragmented data, complex monitoring tasks, traceability requirements, and greenwashing risks. This paper develops a conceptual framework for applying large language models (LLMs) to green power life-cycle accounting and monitoring. Unlike task-specific AI models, the framework positions LLMs as a cross-stage coordination layer that integrates multi-source data management, digital twins, multimodal fusion, edge-cloud deployment, and carbon-footprint tracking. It shows that LLMs can support resource assessment, forecasting, equipment diagnosis, green power traceability, market optimisation, and carbon accounting through a closed loop of perception, modelling, analysis, decision-making, and evolution. It also identifies challenges related to data governance, interpretability, computing emissions, and regulatory accountability, and proposes policy recommendations for data sharing, power-domain models, green computing, and AI-enabled carbon-accounting standards.