红水河2026，Vol.45Issue(2)：25-28,4.DOI:10.3969/j.issn.1001-408X.2026.02.005
基于LightGBM的大藤峡水利枢纽入库流量预测研究
Inflow Prediction of Datengxia Water Control Project Based on LightGBM
丘仕能 1陈晓彬1
作者信息
- 1. 广西大藤峡水利枢纽开发有限责任公司,广西 南宁 530299
- 折叠
摘要
Abstract
To address the insufficient accuracy of traditional inflow forecasting methods for the Datengxia Water Control Project and improve the precision of 24-hour inflow prediction,this study constructs time-series features including historical inflow,upstream hydrological conditions and rainfall based on multi-source heterogeneous data from the Datengxia Data Center.A total of 24 independent LightGBM models are established to fit the mapping relationships between the above features and the hourly inflow at each moment in the subsequent 24 hours.The verification results show that the mean relative prediction error of the proposed model is 17.54％,which is significantly better than the 32.59％of the rainfall and hydrological monitoring system,with an accuracy improvement of 15.05 percentage points.The LightGBM model can effectively characterize the nonlinear coupling relationships among multiple influencing factors and delivers higher forecasting accuracy,which can provide technical support for flood control and power generation dispatching of the water control project.关键词
大藤峡水利枢纽/入库流量预测/LightGBM/时序特征/预测精度Key words
Datengxia Water Control Project/inflow prediction/LightGBM/time-series features/prediction accuracy分类
建筑与水利引用本文复制引用
丘仕能,陈晓彬..基于LightGBM的大藤峡水利枢纽入库流量预测研究[J].红水河,2026,45(2):25-28,4.
