Research on the Construction Technology of a Knowledge Base for Large Models in the Power Industry

Traditional methods for constructing a knowledge base for large models in the power industry primarily rely on general-purpose corpora and simple data augmentation techniques, which struggle to meet the in-depth requirements of the power industry's specialized knowledge. This study significantly enhances the model's ability to understand and apply knowledge in the power domain by constructing a dedicated knowledge graph for the power industry and introducing a chain-of-thought injection mechanism. Experimental results show that, compared to traditional methods, the innovative techniques proposed in this study improve accuracy by 15.3% in power industry knowledge question-answering tasks and increase the F1 score by 12.7% in fault diagnosis tasks. These achievements provide crucial technical support for the construction and application of large models in the power industry.