基于Kriging代理模型的葛洲坝电站水头预测研究  

Research on water head prediction of Gezhouba Hydropower Station based on Kriging surrogate model

作　　者：程晓东 黄斌 赵辉[1] 徐杨[1,2] CHENG Xiaodong;HUANG Bin;ZHAO Hui;XU Yang(Three Gorges Cascade Dispatch and Communication Center,China Yangtze Power Co.,Ltd.,Yichang 443000,China;Hubei Key Laboratory of Intelligent Yangtze River and Hydroelectric Science,China Yangtze Power Co.,Ltd.,Yichang 443000,China)

机构地区：[1]中国长江电力股份有限公司三峡水利枢纽梯级调度通信中心,湖北宜昌443000 [2]中国长江电力股份有限公司智慧长江与水电科学湖北省重点实验室,湖北宜昌443000

出　　处：《人民长江》2023年第6期220-225,共6页Yangtze River

基　　金：国家重点研发计划项目(2019YFC0409000)。

摘　　要：在水电站的调度运行中水头是一项重要数据指标,对机组安全稳定运行起着关键作用。选取非弃水期与弃水期,分析葛洲坝电站不同运行方式下出力分配与弃水流量对大、二江电站水头的影响;引入Kriging代理模型,建立两种典型期下的水头预测模型,并将基于Kriging代理模型得到的水头预测值与实际计算值进行比较分析。结果表明:在非弃水期,大、二江电站水头平均偏差分别为0.22 m和0.26 m,弃水期水头平均偏差分别为0.20 m和0.16 m,预测精度满足实时调度需求。该模型可为水电站水头的趋势变化规律提供一种分析方法,进而为实时优化调度提供决策参考。In the dispatching and operation of hydropower stations,water head is an important data index,which plays a key role in the safe and stable operation of units.After selecting non water abandonment period and water abandonment period,under different operation modes of Gezhouba Hydropower Station,the influence of output distribution and abandoned water discharge on the water head of Dajiang and Erjiang hydropower stations was analyzed.The Kriging surrogate model was introduced to establish a head prediction models in two typical periods.The predicted value of water head based on Kriging surrogate model was compared with the actual calculated value.The result shows that in the non water abandonment period,the average head deviations of Dajiang and Erjiang hydropower stations are 0.22 m and 0.26 m respectively,and in the abandonment period,their average head deviations are 0.20 m and 0.16 m respectively,namely the prediction accuracy meets the needs of real-time dispatching.The model aims to provide a new analysis method for the trend change law of hydropower station s water head,and then provide decision-making reference for real-time optimal dispatching.

关 键 词：水头预测 水电站调度 Kriging代理模型 大、二江电站 葛洲坝电站 

分 类 号：TV697[水利工程—水利水电工程]