Abstract: The traditional power professional knowledge question and answer system faces challenges such as strong professionalism, fast knowledge update, and complex fields, while the existing methods often rely on a single model, resulting in insufficient accuracy and real-time performance. In order to solve this problem, this study proposes a Q&A framework for LLM power expertise that integrates multiple models, and provides more accurate answers by combining the intent recognition model, the text steering model, the knowledge model, and the fine-tuning LLM. The experimental results show that the method based on multi-model fusion and LLM fine-tuning has the highest accuracy and more comprehensive answers in the application of power professional knowledge Q&A than that of only using large models or large models combined with knowledge bases.