融合知识图谱与DeepSeek的抽蓄电站设备运维问答系统研究OA

[Objective]This study develops an intelligent question-answering system that integrates a large language model with a domain-specific knowledge graph to address challenges such as inefficient domain knowledge retrieval,weak decision support,and insufficient professional accuracy of large language models.This system provides operation and maintenance per-sonnel with real-time,precise,and scientifically standardized auxiliary decision support.[Methods]First,a knowledge graph applicable to the operation and maintenance of pumped storage power station equipment was constructed,enabling a systematic representation of equipment operation and maintenance knowledge for the power station.Second,a two-stage knowledge retrieval method was innovatively proposed,deeply integrating the knowledge graph with the DeepSeek large lan-guage model to design and implement the KG-DS QAS,a question-answering system for pumped storage power station equip-ment operation and maintenance.[Results](1)KG-DS QAS achieved precision,recall,and F1 scores of 0.85,0.87,and 0.86,respectively,on the BERTScore evaluation indicator,demonstrating outstanding stability and accuracy.(2)The aver-age score in the expert subjective evaluation was 4.65,and the overall performance significantly outperformed benchmark models(Model Q,Model D,Model L)of the same parameter scale(7B).(3)In timeliness verification,the system's aver-age processing time was 12 seconds,meeting the real-time requirements for on-site operation and maintenance.[Conclusions]KG-DS QAS achieves precise adaptation and application in the field of equipment operation and maintenance for pumped stor-age power stations,providing insights and demonstration for the application of knowledge graph technology and large language models in this domain.Future development can extend from equipment operation and maintenance to full-lifecycle manage-ment of power stations,covering planning,construction,and operation,to build a multidimensional and multimodal integrat-ed intelligent management system.