预测城市用水量的人工神经网络模型研究

Study of artificial neural network model for forecasting urban water demand

为了提高多层前馈神经网络权的学习效率,引入变尺度方法来训练神经网络的权值,并根据训练误差自适应调整学习系数和动量因子.将该方法应用于城市用水量预测中,建立了非线性人工神经网络预测模型.该模型考虑了城市工业用水重复利用率、用水人口、经济发展等众多因素对用水量需求的影响,具备系统决策功能.杭州市预测实例表明,建立的模型及其相应计算方法具有较高的预测精度和准确度.

In order to improve the learning efficiency of multi-layer feedforward neural network, a variable-schedule method was introduced to train the weights of neural network. According to the training errors, the learning coefficients and the momentum factors were self-adaptively adjusted. Applying this method to forecast urban water demand, a forecasting model of nonlinear artificial neural network was established, which included the influences of many factors to urban water demand, such as repeated utilization ratio of industrial water, population, economic development and so on. The proposed forecasting model has systematic decision-making function. The application examples in Hangzhou city show that the forecasting model and its corresponding calculating method has higher forecasting precision and accuracy.