This research introduces a feedback-driven multi-agent framework that transforms LLMs into effective assistants for complex power system simulations.

• Addresses key LLM limitations in power systems: domain knowledge gaps, reasoning constraints, and parameter handling issues
• Positions AI as a versatile research assistant rather than just a problem-solving tool
• Provides a systematic approach to integrate experimental simulations with language models
• Enables more efficient and accurate power system research and engineering workflows

This innovation matters because it bridges critical gaps between AI capabilities and specialized engineering domains, potentially accelerating power grid research, design optimization, and operational improvements.

Enhancing LLMs for Power System Simulations: A Feedback-driven Multi-agent Framework