中国照明电器

通用大模型均基于公开数据进行训练，对道路照明这种细分行业，基于本地数据库实时数据的检索、分析效果并不理想。本研究聚焦于利用大模型技术提升道路照明数据检索能力，针对通用大模型在道路照明行业私有数据检索中的局限性，提出基于数据时效性分类的解决方案。本研究将照明数据分为非时效性数据（如行业标准）和时效性数据（如实时报警记录）两类，并分别设计知识库检索、定时任务预处理、API模板及SQL双重验证等方法，通过时效性数据分类与检索增强生成（RAG）技术结合，使数据检索的准确性提升20%以上。本研究旨在为智慧照明管理系统设计及类似行业的大模型应用提供设计思路和参考。

General large language models are trained on public data,and their performance in data retrieval and analysis for segmented industries such as road lighting,which relies on real-time data from local databases,is not ideal.This study focuses on enhancing road lighting data retrieval capabilities utilizing large language model technology.To overcome the limitations of general large language models in accessing proprietary data within the road lighting industry,it proposes a time-sensitive data classification solution.In this research,lighting data is divided into two categories:non-time-sensitive data(such as industry standards) and time-sensitive data(such as real-time alarm records).Corresponding methods including knowledge base retrieval,scheduled task preprocessing,API templates,and SQL dual verification are designed respectively,with the time-sensitive classification and Retrieval-Augmented Generation(RAG) technology,the system achieves more than 20% improvement in data retrieval accuracy.This study provides a technical reference for the design of smart lighting management systems and demonstrates the transferability to analogous industrial domains.