城市日用水量预测模型比较研究  

Comparing Study on Urban Daily Water Consumption Forecasting Model

作　　者：蒋白懿[1] 代进[1] 高金良[2] 

机构地区：[1]沈阳建筑大学市政与环境工程学院,辽宁沈阳110168 [2]哈尔滨工业大学市政环境工程学院,黑龙江哈尔滨150090

出　　处：《沈阳建筑大学学报（自然科学版）》2008年第2期278-281,共4页Journal of Shenyang Jianzhu University：Natural Science

基　　金：黑龙江省自然科学基金项目(ZJG0503)

摘　　要：目的为保障预测结果准确可靠,建立精度高、可靠性强的城市日用水量模型.方法应用单指数平滑法、灰色方法与BP神经网络方法,分别建立相应日用水量预测模型,并以哈尔滨市的日用水量数据为原始数据进行了实际预测.结果单指数平滑法与BP神经网络模型预测精度较高.用灰色模型预测,所得数值呈递减趋势,其预测精度最低.结论BP网络预测模型是最有效的日用水量预测模型.如果日用水量变化不大,还可采用单指数平滑法预测日用水量.The result of daily forecast for water consumption is very significant for the water supply systems＇ optimum control, and the key to accurate forecast is to find out scientific and reasonable forecast model . Based on the daily water demand data, this paper took exponential smoothing model, grey model and BP neural network model, and got the result of forecasting of daily water consumption. Comparing the result of forecast value and real value, it illustrated that the BP neural network model and the exponential smooth model are available models, however, the result of the Grey model is not satisfied. Therefore, for the shortterm water consumption forecasting, BP neural network model is the most available model. If the water demand almost doesn＇ t change, the exponential smooth model is also suitable for the daily water consumption forecasting.

关 键 词：城市用水量 短期预测 灰色模型 BP神经网络 单指数平滑法 

分 类 号：TU991.33[建筑科学—市政工程]