根据城市时段用水量序列季节性、趋势性及随机扰动性的特点 ,利用人工神经网络方法 ,建立了时间水量短期预报模型 .选取不同的隐层结点数 ,采用相同的输入样本及预测数据进行训练和预测 ,并通过比较其相对误差的大小 ,确定了神经网络的结构 ,并应用 Matlab语言进行了具体的建模和预报 .实例考核证明 ,该方法与常用的时间序列三角函数分析法相比 ,具有预测误差小、计算速度快等特点 ,可满足供水系统调度运行的实际需要 According to the characteristics of seasonal, trend and stochastic perturbations of urban water use sequence, a short-term forecasting model of time and water is established by using artificial neural network method.Selecting different hidden layer nodes and using the same input samples and forecasting data for training And predictions, and by comparing the relative error of the size of the neural network to determine the structure, and the use of Matlab language for a specific modeling and forecasting.Example assessment shows that the method compared with the commonly used time series trigonometric function analysis, with Small prediction error, fast calculation speed, etc., to meet the actual needs of water supply system scheduling operation