BP和RBF神经网络在水轮机非线性特性拟合中的应用比较  

Application of BP Neural Network and RBF Neural Network in Extending Hydraulic Turbine Combined Characteristic Curve

作　　者：张培[1] 陈光大[1] 张旭[1] 

机构地区：[1]武汉大学动力与机械学院,武汉430072

出　　处：《中国农村水利水电》2011年第11期125-128,131,共5页China Rural Water and Hydropower

摘　　要：利用神经网络对水轮机综合特性曲线进行数据处理和延伸,不必建立具体的函数关系表达式,就可对已知的离散数据进行拟合。并且还可以结合边界约束条件对未知区域内的数据进行预测,从而提高了水轮机综合特性曲线数据处理的工作效率和数据精度。分别介绍了用BP神经网络和RBF神经网络对水轮机综合特性曲线数据处理和延伸的方法。并采用一机组的样本数据进行训练,比较2种方法的训练结果得出结论。It is unnecessary to establish concrete function expression, the known discrete data can be fitted by using neural network to extend hydraulic turbine combined characteristic cure. And we can also add boundary conditions to predict unknown zones, so as to raise the work efficiency and data precision in data treatment concerning hydraulic turbine combined characteristics. This paper intro- duces the use of gP neural network and RBF neural network in extending hydraulic turbine combined characteristic curve. I.astly, the results of the two methods are compared and some conclusions are obtained.

关 键 词：水轮机综合特性曲线 BP神经网络 径向基神经网络 曲面拟合和延伸 

分 类 号：TK730.2[交通运输工程—轮机工程]