小波神经网络改进算法在来水量预测中的应用  

Application of WNN Improved Algorithm in Inflow Water Forecast

作　　者：陶猛[1] 徐淑琴[1] 李洪涛[1] 

机构地区：[1]东北农业大学水利与建筑学院,黑龙江哈尔滨150030

出　　处：《节水灌溉》2013年第10期38-40,共3页Water Saving Irrigation

基　　金：黑龙江省教育厅科研项目<黑龙江省灌区水资源动态预测方法与应用研究>(11551044)

摘　　要：中小河流的径流量表现出十分复杂的变化特征,如高度非线性、多时间尺度特征、时频序列,严重影响预测的准确性。由于常规的分析法难以建立有效的预测模型,所以为了提高预测精度,提出了一种改进小波神经网络的来水量预测模型,利用非线性小波函数取代了BP神经网络通常用的Sigmoid函数作为隐含层节点的传递函数,有效地避免神经网络结构设计的盲目性,同时也有更强的学习能力且精确更高。取实例建模分析,并建立BP网络模型与之比较,结果表明,小波神经网络提高了径流量预测精度。The series of runoff in small and medium-sized river display complex features, such as highly non-linear, multi-time scale features which change with the time. This seriously affects the prediction accuracy. Since it is difficult to establish a model based on the regular analysis methods, in order to improve the accuracy, an improved Wavelet-Neural Network model is put forward to predict the water, which uses a non-linear wavelet function as the transfer function of the hidden layer nodes instead of the regular non-linear -Sigmoid function in BP Neural Network and avoids the blindness of the structure design of the Neural Network and get better ability to learn more accurately. Then a model for real-case is established and compared with the BP Neural Network. The result shows that using the improved Wavelet-Neural Network to predict the inflow water is more accurate.

关 键 词：中小河流径流量 SIGMOID函数 小波神经网络 BP神经网络 

分 类 号：TV121[水利工程—水文学及水资源]