城市供水DMA短期需水量预测比较研究

Comparative Research on Short Term Water Demand Prediction of Urban Water Distribution System

针对城市供水独立计量区域(DMA)实例,运用自回归积分滑动平均模型、Elman神经网络、广义回归神经网络以及最小二乘支持向量机模型进行城市供水管网短期需水量预测.结合节假日、天气状况及温度等因素对用水量的影响,修正了相应模型,进一步提高了需水量预测精度.实例计算结果表明,4种模型均能用于城市供水短期需水量预测,其中结合天气及节假日因素的最小二乘支持向量机模型具有更高的预测精度.

Applied four methods including the ARIMA, ELMAN, GRNN and LSSVM to predict short-term water demand for a DMA instance. Considering the impact of water demand by weather conditions, temperature and holidays and other factors, so combined these factors can improve water demand forecast accuracy. The final results show that all four methods can be used for short-term water demand prediction, and the LSSVM method, which combined with weather and holiday factors, has higher prediction accuracy.