基于神经网络方法和支持向量机的西安市水资源需水预测

Forecast of water resource demand in Xi'an city using neural network and support vector machine models

【目的】建立精度更高的需水量预测模型,为水资源规划管理提供理论依据。【方法】建立基于神经网络方法和支持向量机的需水量预测模型,以西安市需水预测为例,对2种预测模型的预测结果进行了比较。【结果】利用建立的径向基函数神经网络需水预测模型,得到西安市2010和2020年的需水量分别为32 485.65,48 180.43万m3;采用支持向量机模型能全面考虑影响需水预测的各种因素,预测西安市2010和2020年的需水量分别为32 488.03,48 184.41万m3。【结论】基于神经网络方法和支持向量机方法的需水量预测模型均可全面反映需水量的变化规律,两者预测结果十分接近,均可用于需水量的精确预测。

【Objective】 Water demand forecast models with higher precision were established to improve water resources planning.【Method】 This study established neural network and support vector machine based forecasting models and used them to predict water demand in Xi＇an.【Result】 The predicted water demands of Xi＇an in 2010 and 2020 by the established radial basis function neural networks forecasting model were 324 856.5 thousand m3 and 481 804.3 thousand m3,respectively.Support vector machine model,which could fully consider the various factors affecting water demand forecasts,predicted the water demands in 2010 and 2020 were 324 880.3 thousand m3 and 481 844.1 thousand m3,respectively.【Conclusion】 Both of the two built models based on neural network and support vector machine were able to accurately predict water demand.