地下河日流量预测模型  

Prediction Model of Underground River Daily Discharge

作　　者：束龙仓[1] 陶玉飞[1] 鲁程鹏[1] 董贵明[1] 王茂枚[1] 刘丽红[1] 

机构地区：[1]河海大学水文水资源与水利工程科学国家重点实验室,江苏南京210098

出　　处：《水电能源科学》2008年第3期1-3,共3页Water Resources and Power

基　　金：国家重点基础研究发展计划(973项目)(2006CB403200);高等学校学科创新引智计划基金资助项目(B08048);教育部长江学者和创新团队发展计划基金资助项目(IRT0717)

摘　　要：将门限自回归模型(TAR)和小波—BP神经网络组合模型应用于后寨河流域出口流量的预测中,建立了阶数分别为5、4、1三段的自回归模型。采用db3小波对地下河日流量序列进行分解作为BP神经网络的输入,建立小波—BP神经网络组合模型。从绝对误差和相对误差角度对比分析两种模型,得出小波—BP神经网络组合模型更适合本地下河日流量预测。针对两模型的不足提出了改进的建议。This paper introduced threshold auto-regressive model and combined prediction model of wavelet and BP neural network, and the outlet discharge of Houzhai River basin was predicted with these two models. The threshold auto-regressive model included three auto-regressive models with orders of 5, 4, and 1 respectively. The daily discharge series were decomposed by db3 wavelet used as the input of BP neural network model, and the combined prediction model of wavelet and BP neural network was established. This paper compared absolute error and relative error of these two models, and the combined prediction model of wavelet and BP neural network was better. At last, according to the shortcomings of the two models, some improvement suggestions are proposed.

关 键 词：地下河日流量 门限自回归 小波 BP神经网络 

分 类 号：TV211.12[水利工程—水文学及水资源]