水电站变出力系数的神经网络估计方法

Neural network estimation methods for varying output coefficients of hydropower stations

围绕如何提高水电站中长期发电调度出力系数估计和出力过程计算精度的问题,综合考虑水电站水头、发电流量等水电站关键状态信息对综合出力系数K的影响,建立了以水头、发电流量和入库流量为备选输入,以综合出力系数K为输出的三种神经网络模型,进而提出了水电站发电调度出力计算变K值的三种神经网络估计方法。结合三峡水电站多年实际运行资料,将本文提出的变K值估计方法与多种传统K值确定方法进行了综合对比。结果表明,本文提出的变K值估计模型或方法具有更高的K值估计、出力和发电量计算精度,为实现水电站中长期发电调度精细化出力计算提供了新途径,具有显著的工程应用价值。

Focusing on how to improve the accuracy in estimating power output coefficients and power generation process of a hydropower station for its medium-long-term generation dispatching,this study develops three neural network models with an output of power output coefficient K and three alternative inputs–water head,generation flow,and inflow rate.And we formulate three neural network methods for estimating the varying K values in the station’s output calculations and compare the methods with traditional methods,in a case study of the many years’operation data of the Three Gorges hydropower station.The results show our methods are more accurate in the estimations of K value,power output,and power generation,providing a new practical approach to the refined output calculation of medium-longterm power generation dispatching of hydropower stations.