BP网络的时用水量预测组合模型城市.pdfVIP

4 1 6 V ol4 1 N o6 2009 6 JOU RNAL OF HARB IN IN ST ITUTE OF TECHNOLOGY Jun. 2009 BP 1 2 3 陈 卫, 陆 健 , 吴志成 1. , 2 10098, cw 5 826@ hhu. edu. cn; 2. , 21000 3; 3. , 210002) : 为建立起城市用水量与其影响因素间的预测模型, 以预测的城市用水量趋于合理, 针对城市时用水 量的特点 影响因素, 在考虑充分利用各因素历史观测数据的基础上, 利用 BP神经网络建立了城市时用水 量的时间序列预测与解释性预测组合模型, 并对南京市的时用水量进行了预测. 预测结果与实际情况具有很 好的一致性, 预测误差小, 能满足供水系统调度的实际需要. 可见, 本预测组合模型是合理的, 为城市时用水 量预测提供了 一种可行方法. : 时用水量; 组合模型; 预测; BP 神经网络 : TU 99 13 : A : 0 367 - 6234 2009) 06- 0197 - 04 Com b ined forecast m odel of urban hourly water consum p tion based on BP neural n etw ork 1 2 3 CHEN W e i, LU J ian , WU Zh icheng 1. Co llege of Env ironm enta l Sc ience and Eng ineer ing, H ohai U nivers ity, N anj ing 21009 8, Ch ina, cw 5826@ hhu. edu. cn; 2. J iansu Eng ineer ing Consu lting Cen ter, N an jing 21000 3, China; 3. N an jing W ater Supply G enera l Com pany, N an j ing 210002, Ch ina) A bstract: T o establish the re lationsh ip betw een urban w ater consum ption and its affect ing factors, according to the characterist ics of urban hourly w ater consum pt ion, a com bined urban hourly wa ter consum ption pred iction m odel has been deve loped based on BP neura l netwo rk. The m odel has been perform ed on the h istorica l obser vation data o f N an jing city. T he results show that the fo recast error o f the developed m ode l is sm all and m eets the practical requ irem ent o f w atersupply dispatch system. It can be conc luded that this mode l is reasonab le and feasib le for the fo recast of urban hourly w