人民黄河2025，Vol.47Issue(11)：60-64,70,6.DOI:10.3969/j.issn.1000-1379.2025.11.009
基于梯度提升的凌情特征日期预报模型研究
Gradient Boosting-Based Forecasting Model for Characteristic Dates of Ice Regime
摘要
Abstract
Accurate forecasting characteristic dates of ice regime is an extremely important aspect of winter ice disasters prevention and dis-patching work in reservoirs.In order to meet the demands of winter ice prevention regulation work at the Wanjiazhai Reservoir,classification model and regression model for the Toudaoguai hydrological station section based on the gradient boosting algorithm were developed.The per-formances of the two types of models were compared and evaluated,and the SHAP interpreter was used to analyze the test results.The results show that the test accuracy of the ice floe date,river closure date and river opening date of the classification model all achieves Grade A,which effectively solves the problem of insufficient model accuracy caused by the lack of training samples.Furthermore,the forecast accuracy of the classification model is significantly better than that of the regression model,suggesting that the prediction of characteristic dates of ice regime can be approached as a classification issue.Air temperature is the most important factor contributing to the prediction of river icing characteristic dates,with a contribution of over 70%.关键词
凌情特征日期/梯度提升算法/分类模型/回归模型/SHAP/万家寨水库Key words
characteristic dates of ice regime/gradient boosting algorithm/classification model/regression model/SHAP/Wanjiazhai Res-ervoir分类
资源环境引用本文复制引用
陈清闲,孙超,脱友才,路新川,李嘉,黄文典..基于梯度提升的凌情特征日期预报模型研究[J].人民黄河,2025,47(11):60-64,70,6.基金项目
国家重点研发计划项目(2022YFC3202500) （2022YFC3202500）
