城市供水管网水量预测小波神经网络方法.pdf

38 7 Vo.l 38 No. 7 2005 7 Journal of Tianjin University Ju.l 2005 * 刘洪波, 张宏伟, 闫晓强 (, 300072) : 为提高城市供水优化调度的可靠性和实用性, 对城市管网水量预测的方法进行了研究. 提 了利用小波 分解与人工神经网络相结合的小波神经网络管网水量预测模型, 该模型以非线性小波基为神经元变换函数, 通过 伸缩因子和平移因子计算小波基函数合成的小波网络, 并从理论上给 了严密的算法; 同时通过逐步检验算法, 科 学地确定了网络结构, 克服了普通人工神经网络难以确定网络结构存在局部极小点等缺点. 仿真结果表明, 该模 型比普通人工神经网络预测模型的预测精度高, 并具有很强的适应能力. : ; ; : TU991. 31 : A : 2005) WaveletNeutralNetwork ForecastingMethod forWaterConsumption inMunicipal SupplyWaterNetworks LIU Hong o, ZHANG Hongw e,i YAN X iaoqiang ( School of Environment Science and Engineering, T ianjin University, T ianjin 300072, Ch ina) bstract: In order to mi prove the relia ility and the practica ility of optmi al operation ofw ater supply system, a forecastingm ethod forwater consum ption in m un icipal supplyw ater newt orks is put forward. Th ism ethod is ased on wavelet neutralnetwork in which the nonlinearw avelet asis function is used as the transform function of neurons instead of sigm oid function. Thewavelet network can e got y calculating the flex ing and the expan sion factors. Moreover, a rigorous arithmetic is proposed in theory. And also the network configuration can e scientifical confirm ed through checking the arithm etic step y step. Compared w ith the ANN method, th is method is easy in configuration decision and can overcom e the localm in mi al pro lem. S mi ulation results show that thisw avelet neutral network forecastingm ethod has the advantages of high forecasting accuracy and strong applica ility. Keywords: m unicipal supplyw ater networks; water