Enhancing Auxiliary Decision-Making in Power Supply Service Command via a Text-Based Large Model Framework

With the rapid transformation of modern power grids, traditional decision-support tools have increasingly struggled to keep pace with dynamic operational demands. Addressing this gap, we introduce PowerCMD, a decision-support system explicitly tailored to the power supply service command domain. Unlike existing general-purpose frameworks, PowerCMD uniquely combines a large-scale generative language model with operational power-grid data and structured domain-specific knowledge. The core innovation lies in its adaptive prompting method, enabling context-aware reasoning across various realtime operational scenarios, including equipment diagnostics, proactive maintenance, and customer incident handling. Experimental validation using real-world operational datasets reveals that PowerCMD substantially improves both the accuracy of recommendations and the responsiveness required by frontline operational teams, demonstrating clear practical benefits over baseline models.