AI Agents in Power System Operation: Application Results and Future Potentials

The increasing complexity of distribution grid management demands innovative solutions for system operators. This paper presents a novel approach for a distribution grid operator assistants using Large Language Models. We implement an LLM agent that autonomously analyzes grid states, identifies problems, and implements corrective measures to maintain grid stability. The agent leverages the ReAct (Reasoning and Acting) framework to combine reasoning capabilities with power system analysis tools to support distribution grid management. First experiments using simulation models demonstrates the agent's ability to consistently resolve line loading issues through grid topology reconfigurations or transformer tap adjustments. The results show that the agent successfully addresses thermal constraints, voltage rise, and reactive power challenges across multiple operational scenarios with high consistency in decision-making. The approach offers a promising solution to the increasing complexity of distribution grid management in the context of renewable energy integration and sector electrification while maintaining transparency in the decision-making process.