水利学报2025，Vol.56Issue(9)：1166-1177,12.DOI:10.13243/j.cnki.slxb.20250042
基于大语言模型的水库群防洪调度规则标准化及特征研究
Standardization of reservoir groups operating rules and analysis of their characteristics using large language models
摘要
Abstract
Reservoir operating rules are often described in natural language,outlining operating principles,objec-tives,and constraints.However,these descriptions typically suffer from inconsistencies in format,terminological dis-crepancies,and a lack of clear structure,which present significant challenges for the standardization and digital pro-cessing of these rules.This study focuses on the controlled reservoir groups in the Yangtze River Basin and utilizes a locally deployed large language model to standardize the text-based flood control operating rules into well-structured decision tree representations.A method to measure the complexity of these operating rules was proposed by analyzing decision tree depth and nodes,thereby uncovering the relationship between key decision variables,their complexity,and the characteristics of reservoir groups.The results indicate that decision tree-based operating rules provide a more intuitive and user-friendly framework for operating,in contrast to traditional text descriptions.Significant differ-ences in operating rule complexity were observed across different reservoirs,though the correlation with most charac-teristic variables of the reservoir groups was found to be low.While spatial clustering patterns in reservoir group fea-tures were identified,their connection to decision complexity remained weak.The joint operating plan for the con-trolled reservoir groups in the Yangtze River Basin should prioritize top-level design to minimize inconsistencies in operating rules across reservoirs,thus advancing the standardization and digitization of these rules.关键词
水库群/联合调度/大语言模型/决策树/特征分析Key words
reservoir groups/joint operation/large language model/decision tree/feature analysis分类
建筑与水利引用本文复制引用
任康,张睿,黄强..基于大语言模型的水库群防洪调度规则标准化及特征研究[J].水利学报,2025,56(9):1166-1177,12.基金项目
国家自然科学基金项目(52309025) （52309025）
