城市需水的智能预测方法研究

选取对城市用水影响较大的因素作为预测指标,通过HP滤波分析其指标及用水量的趋势成分及波动成分。应用多元线性回归法对趋势性成分进行模拟,应用模糊神经网络对波动性成分进行网络训练,以获得城市需水的总预测值。以大连市1980~2000年用水及其相关因子为例建模,以2001~2007年指标对模型进行检验,预测了大连市2010、2020年的需水值,为城市水资源规划提供一定的理论支持。

Study on method for intelligent forecasting of urban water demand

The factors of large impact on the urban water use are taken as the forecasting indexes，and then the indexes and the components of both the trend and fluctuation of the water consumption are analyzed through Hodrick Prescott filter. Furthermore，the trend component is simulated with the method of the multivariable linear regression，while the network training is made on the fluctuation component with the fuzzy neural network，so as to obtain the gross forecasting value. By taking the water consumption of 1980~2000 of Dalian and its relevant factors as the case for modeling and verifying the model with the indexes from 2001 to 2007，the water demand values of both 2010 and 2020 are forecasted for providing a certain theoretical support for the urban water resources planning concerned.