基于机器学习的电站耗水率计算及其对水位预测影响研究  

Machine Learning-based Simulation of Water Consumption Rate of Hydropower Station and Its Impact on Water Level Forecast

作　　者：张森 谢帅 王永强[1,2] 徐杨 肖蕴珂 ZHANG Sen;XIE Shuai;WANG Yongqiang;XU Yang;XIAO Yunke(Water Resources Department,Changjiang River Scientific Research Institute,Wuhan 430010,Hubei,China;Hubei Key Laboratory of Water Resources&Eco-Environmental Sciences,Wuhan 430010,Hubei,China;Three Gorges Cascade Dispatch&Communication Center,Yichang 443002,Hubei,China)

机构地区：[1]长江科学院水资源综合利用研究所,湖北武汉430010 [2]流域水资源与生态环境科学湖北省重点实验室,湖北武汉430010 [3]三峡水利枢纽梯级调度通信中心,湖北宜昌443002

出　　处：《水力发电》2023年第12期94-100,共7页Water Power

基　　金：国家重点研发计划重点专项(2022YFC3202300);国家自然科学基金面上项目(42271044);中央级公益性科研院所基本科研业务费项目(CKSF2021486);数字孪生长江水工程智能调度关键技术研究项目(SKS2022128)。

摘　　要：耗水率是水电站坝前水位、尾水位预测的主要影响因素,也是电站发电计划编制和经济运行的基础,其精确计算至关重要。以三峡水库为研究对象,提出了基于机器学习的耗水率精细化计算方法。构建了基于水量平衡的水位预测模型,借助机器学习算法对耗水率与尾水位进行精确计算,并以此为基础预测2 h坝前水位与尾水位,观察机器学习算法应用对水位预测精度的提升。结果表明,在不同运行场景下,所提方法对于提升电站耗水率值、坝前和尾水位预测精度效果较好,电站耗水率误差绝对值降低60%以上,坝前水位预测误差降低约12%,尾水位计算误差降低约88%,可为电站安全稳定经济运行提供参考与支持。Water consumption rate is the main influencing factor for forecasting upstream water level and tail water level of hydropower station and is also the basis of power generation planning and economic operation of hydropower station,so its precise calculation is crucial.Taking Three Gorges Reservoir as the research object,a refined calculation method of water consumption rate based on machine learning is proposed.A water level forecast model based on water balance is constructed,and the water consumption rate and tail water level are accurately calculated by machine learning algorithm.On this basis,the upstream water level and the tail water level within 2 hours are predicted,and the improvement of water level prediction accuracy with the application of machine learning algorithm is observed.The results show that,under different operation scenarios,the proposed method has a good improvement effect on the forecast accuracy of water consumption rate and water levels,with a decrease of over 60%in the MAE of water consumption rate,a decrease of about 12%in the prediction error of upstream water level and a decrease of about 88%in the calculation error of tail water level,which can provide reference and support for the safe,stable and economic operation of power station.

关 键 词：耗水率拟合 坝前水位 尾水位 机器学习 预测 三峡水库 

分 类 号：P731.34[天文地球—海洋科学]