基于MEA-BP的水轮机运转特性曲线拟合及应用

Fitting and application of hydraulic turbine operating performance curve based on MEA-BP neural network

针对目前水轮机综合特性曲线数据拟合精度的局限性以及水电站运行过程中水轮机参数特性掌握应用不足的问题,利用图形数据提取软件,经数据转换与计算,将获取到的运转特性曲线正常运行区的水头、出力、效率和流量等基本工作参数作为样本数据,分别应用具有非线性逼近能力的MEA-BP神经网络和传统的BP神经网络对样本进行网络训练和测试,并将测试结果与样本数据进行了对比分析,同时还对误差进行了计算分析。结果表明:两种拟合方法均有效,而且MEA-BP方法的拟合精度优于传统BP方法的拟合精度;利用MEA-BP方法对运转特性曲线进行有效拟合,可为智能水电厂智能调度的自动控制系统设计提供技术支撑;同时也有助于电站运行人员更直观地预判电站机组的运行工作状态,指导电站高效运行。

In the light of accuracy limitation in data fitting of hydraulic turbine operating performance curve and insufficient mastery and application for turbine parameter characteristics in the operation of hydraulic power station,by the image data extraction software and through data transfer and calculation,the basic working parameters of water head,power output,efficiency and discharge etc.obtained from operating performance curve in normal operating zone were trained and tested by MEA-BP neural network that possesses nonlinear approximation ability and traditional BP neural network.The test results were compared with the sample data and the errors were calculated.The results showed that both the two data fitting methods are effective,and MEA-BP was better than traditional BP in data fitting.The effective fitting of hydraulic turbine operating performance curve by MEA-BP can provide technology support for the automatic control system of hydropower’s intelligent operation.At the same time,it is helpful for the staffs of the power station to prejudge the working state of the hydraulic turbine more intuitively and keep the hydropower station operating more efficiently.