摘要
【目的】将小波变换与支持向量机结合,构建小波支持向量机回归模型(WSVR),并用其对日径流进行预测,为水库调度提供参考依据。【方法】利用径流时间序列中包含的大量信息,通过小波变换将径流时间序列分解成不同分辨率水平的子序列和近似序列,通过相关性分析选取有效子序列与近似序列相加得到的新序列作为支持向量机回归模型的输入,建立小波支持向量机回归耦合模型,以泾河流域张家山站的日径流为研究对象,利用均方根误差(RMSE)、平均绝对误差(MAE)、确定性系数(DC)、相关系数(R)及相对误差(RE)作为评价指标对模型预测精度进行评价。【结果】利用所建立的小波日径流支持向量机模型对张家山站日径流的预测结果显示,该模型在检验阶段的RMSE、MAE、DC、R及RE分别为26.05m3/s,8.26m3/s,0.826,0.910,-13.3%,与仅使用支持向量机回归模型(SVR)相比,耦合模型预测精度明显提高,且非汛期预测效果优于汛期。【结论】建立了小波支持向量机回归耦合模型,该模型可有效模拟和预测日径流,为日径流预测提供了新的途径。
[Objective]The wavelet support vector machine（WSVR）regression model was established by integrating wavelet with support vector machine for forecasting runoff to provide reference for better reservoir operation.[Method]Original time series were decomposed into subseries of different resolution levels and approximate sequence using wavelet techniques,and effective subseries were finally selected through correlation analysis.Then wavelet support vector machine regression model（WSVR）was constructed using the new series that adding approximate sequence with effective subseries as input.The model was applied to forecast daily runoff at Zhangjiashan Station of Jinghe River,and root mean square error（RMSE）,mean absolute error（MAE）,deterministic coefficient（DC）,correlation coefficient（R）and relative error（RE）were calculated to evaluate the model.[Result]From the daily runoff forecast at Zhangjiashan Station,the established WSVR model had RMSE,MAE,DC,Rand REof 26.05m3/s,8.26m3/s,0.826,0.910,and-13.3%,respectively.The accuracy of coupled model was much higher than thatof support vector machine（SVR）model.The WSVR model was more accurate in flood season than in dry season.[Conclusion]The established WSVR model could be applied to forecast daily runoff effectively,which provided a new way for runoff forecast.
出处
《西北农林科技大学学报（自然科学版）》
CSCD
北大核心
2016年第4期211-217,共7页
Journal of Northwest A&F University(Natural Science Edition)
基金
水利部公益性行业科研专项(201301016)
"十二五"国家科技计划项目(2012BAD08B01)
西北农林科技大学中央高校基本科研业务费科技创新重点项目(QN201168)
关键词
日径流预测
小波变换
支持向量机
张家山水文站
daily runoff
wavelet transforms
support vector machine
Zhangjiashan Station
