| Water-use forecasting is important for municipal construction planning and the operation of water-supply system. Water-use forecasting is composed of the yearly, seasonal, monthly, daily and hourly municipal water-use forecasting. In the several parts, daily one has a key-position. It may not only directly guide the production of water-supply corporations, but also offer the technical service for optimal scheduling among several water-supply corporations. This paper has established nonparametric regression model, partially linear AR model and other forecasting models of xi'an urban daily water demand based on the xi'an daily water-use data. The four aspects are as follows:The multi linear regression model has been set up and the prediction error is checked by theχ2 method. The prediction error is corrected, at last xi'an daily water-use synthesis forecasting model is set up and the prediction error is not good.Based on the regression function estimated by the kernel estimation and local linear estimation, the nonparametric multi regression model of xi'an urban daily water demand is set up. By the comparison of fact date, it was proved that the nonparametric multi regression model can meet the practical requirement of water supply dispatch system.According to BJ time series method, the linear AR model of xi'an water daily demand is set up. The computation result shows that the forecasting model for the daily water-consumption is not good.The partially linear AR model of xi'an water daily demand is set up. The linear aspect takes into account the water demand and the nonlinear aspect takes into account tiptop temperature. This model not only has the merit of regressive model but also has the linear autoregressive model merit, so its forecasting result is better. By the comparison of the former three models, this method is best for the xi'an water daily demand forecasting. |
