基于GRU和XGBoost算法的中小流域流量预测研究  

Research on small and medium-sized watershed flow prediction based on GRU and XGBoost algorithm

作　　者：苏林鹏 周祥[2,3] 张守平 Su Linpeng;Zhou Xiang;Zhang Shouping(Yuxi Design Institute of Water Conservancy Electric Power Co.,Ltd.,Chongqing402160,China;ChongQing Water Resources And Electric Engineering College,Chongqing402160,China;Reservoir Safety and Big Data Environment Chongqing University Engineering Center,Chongqing402160,China)

机构地区：[1]重庆市渝西水利电力勘测设计院有限公司,重庆402160 [2]重庆水利电力职业技术学院,重庆402160 [3]水库安全及水环境大数据重庆市高校工程中心,重庆402160

出　　处：《吉林水利》2024年第10期72-78,共7页Jilin Water Resources

基　　金：重庆市技术创新与应用发展专项重点项目(CSTB2022TIAD-KPX0132)。

摘　　要：中小流域的流量预测对于水资源管理和防洪工程等具有重要意义,因此针对传统小流域流量预测模型精度不高、实时性不强的问题,提出一种基于门控循环神经网络(Gate Recurrent Unit,GRU)和XGBoost的中小流域流量预测方法。在该项研究中,首先利用多传感器协同传输技术采集了流域内的水文数据,如降雨量和径流量,并依据GRU算法捕获数据在时序维度的关联关系。然后,将GRU模型提取的时序特征作为XGBoost模型的输入,利用XGBoost算法实现中小流域流量的预测。最后,在某真实小流域场景中采集流量数据,并根据预测数据与实际观测数据的对比和分析,评估所提出方法的有效性和准确性。实验结果显示,基于GRU和XGBoost算法的中小流域流量预测方法能够较为准确地预测流域的流量变化,为水资源管理和防洪工程提供可靠的决策依据。Forecasting of flow in small and medium-sized watersheds is of great significance for water resources management and flood control projects.To address the problems of low accuracy and poor real-time performance of traditional small watershed flow forecasting models,a method based on Gate Recurrent Unit(GRU)and XGBoost is proposed.In this study,hydrological data such as rainfall and runoff are collected within the watershed using multi-sensor collaborative transmission technology,and the correlation between data in the time dimension is captured using the GRU algorithm.Then,the temporal features extracted by the GRU model are used as input to the XGBoost model,which is used to predict the flow in small and medium-sized watersheds.Finally,flow data are collected from a real small watershed scenario,and the effectiveness and accuracy of the proposed method are evaluated through comparison and analysis of the predicted data with actual observation data.The experimental results showed that the flow forecasting method for small and medium-sized watersheds based on GRU and XGBoost algorithms could accurately predict changes in flow in the watershed,providing reliable decision-making support for water resources management and flood control projects.

关 键 词：中小流域流量预测 门控循环神经网络 XGBoost 多传感器协同传输技术 

分 类 号：TV124[水利工程—水文学及水资源]