基于WNN-SVM的水文时间序列预测方法研究
|
摘要
在面对具有随机性、突变性的复杂时间序列数据(如流量等水文时间序列数据)时,传统单一的模型的预测精度不尽人意,对单一模型的优化不能完全克服其局限性。因此,文中提出一种基于WNN-SVM组合的水文时间序列预测模型。首先对水文时间序列数据作均值归一化处理,然后对预处理后的水文时间序列进行小波分解,将分解后的子序列通过相空间重构的方法使其从低维时间序列向高维转换;根据其分解后的特点,对尺度变换序列采用支持向量机(SVM)进行建模预测,小波变换序列采用小波神经网络(WNN)进行建模预测,再将两者的预测结果进行小波重构,得到原始序列预测值。随后采用屯溪流域1980年至2007年43 996个小时流量数据进行实验验证,结果表明该模型的预测精度高于单一模型,证明了该模型的有效性。
When faced with hydrological time series data with randomness and inconsistency,such as flow,etc.,the prediction accuracy of traditional single model is not satisfactory,and the optimization of single model cannot completely overcome its limitations. Therefore,we present a hydrological time series prediction model based on WNN-SVM combination. First of all,the hydrological time series data are normalized by means,and then the wavelet decomposition is performed on the pretreatment hydrological time series and the low-dimensional time series is transformed to high-dimensional by phase space reconstruction. According to the characteristics of decomposition,the scaling sequence is modeled and predicted by support vector machine(SVM),and the wavelet transform sequence is modeled and predicted by wavelet neural network(WNN),and the predicted value of the original sequence is obtained after the wavelet reconstruction of the predicted result. The experiment of 43 996 data of hourly flow in Tunxi basin from 1980 to 2007 indicates that the prediction accuracy of the model is higher than that of a single model,which proves the validity of the model.
When faced with hydrological time series data with randomness and inconsistency,such as flow,etc.,the prediction accuracy of traditional single model is not satisfactory,and the optimization of single model cannot completely overcome its limitations. Therefore,we present a hydrological time series prediction model based on WNN-SVM combination. First of all,the hydrological time series data are normalized by means,and then the wavelet decomposition is performed on the pretreatment hydrological time series and the low-dimensional time series is transformed to high-dimensional by phase space reconstruction. According to the characteristics of decomposition,the scaling sequence is modeled and predicted by support vector machine(SVM),and the wavelet transform sequence is modeled and predicted by wavelet neural network(WNN),and the predicted value of the original sequence is obtained after the wavelet reconstruction of the predicted result. The experiment of 43 996 data of hourly flow in Tunxi basin from 1980 to 2007 indicates that the prediction accuracy of the model is higher than that of a single model,which proves the validity of the model.
引文
[1] 唐黎哲,冯大为,李东升,等.以LDA为例的大规模分布式机器学习系统分析[J].计算机应用,2017,37(3):628-634.
[2] HAN Jiawei,KAMBER M.Data mining:concepts and techniques[M].2nd ed.[s.l.]:Morgan Kaufmann Publishers,2004.
[3] 周金玉,万定生,肖艳.基于嵌入式索引的水文时间序列预测模型[J].微电子学与计算机,2018,35(1):118-123.
[4] BOX G E P,JENKINS G M,REINSEL G C.Time series analysis:forecasting and control-4t[M].[s.l.]:John Wiley & Sons,Inc.,2008.
[5] KOTTEGODA N T.Stochastic water resources technology[M].London:Macmillan Press,1980.
[6] WEI Shouke,YANG Hong,SONG Jinxi,et al.A waterlet-neural network hybrid modeling approach for estimating and predicting river monthly flows[J].Hydrological Sciences Journal,2013,58(2):374-389.
[7] 成云,成孝刚,谈苗苗,等.基于ARIMA和小波神经网络组合模型的交通流预测[J].计算机技术与发展,2017,27(1):169-172.
[8] SEO Y,KIM S,KISI O,et al.Daily water level forecasting using wavelet decomposition and artificial intelligence techniques[J].Journal of Hydrology,2015,520:224-243.
[9] 刘伟,李春青,张艳芬.基于小波分解和最小二乘支持向量机的COD预测[J].微电子学与计算机,2013,30(4):26-29.
[10] ZHANG Q,BENVENISTE A.Wavelet networks[J].IEEE Transactions on Neural Networks,1992,3(6):889-898.
[11] 杨颖,陈德华.基于小波神经网络的时间序列流数据的研究[J].计算机技术与发展,2006,16(6):193-195.
[12] 鞠琴,余钟波,郝振纯,等.小波网络模型及其在日流量预测中的应用[J].西安建筑科技大学学报:自然科学版,2009,41(1):47-52.
[13] 朱跃龙,李士进,范青松,等.基于小波神经网络的水文时间序列预测[J].山东大学学报:工学版,2011,41(4):119-124.
[14] 李代华,崔东文.相空间重构支持向量机在径流模拟中的应用研究[J].长江科学院院报,2013,30(10):21-26.
[15] 廖荣华,兰时勇,刘正熙.基于混沌时间序列局域法的短时交通流预测[J].计算机技术与发展,2015,25(1):1-5.
[16] FALAMARZI Y,PALIZDAN N,HUANG Y F,et al.Estimating evapotranspiration from temperature and wind speed data using artificial and wavelet neural networks(WNNs)[J].Agricultural Water Management,2014,140(3):26-36.
[2] HAN Jiawei,KAMBER M.Data mining:concepts and techniques[M].2nd ed.[s.l.]:Morgan Kaufmann Publishers,2004.
[3] 周金玉,万定生,肖艳.基于嵌入式索引的水文时间序列预测模型[J].微电子学与计算机,2018,35(1):118-123.
[4] BOX G E P,JENKINS G M,REINSEL G C.Time series analysis:forecasting and control-4t[M].[s.l.]:John Wiley & Sons,Inc.,2008.
[5] KOTTEGODA N T.Stochastic water resources technology[M].London:Macmillan Press,1980.
[6] WEI Shouke,YANG Hong,SONG Jinxi,et al.A waterlet-neural network hybrid modeling approach for estimating and predicting river monthly flows[J].Hydrological Sciences Journal,2013,58(2):374-389.
[7] 成云,成孝刚,谈苗苗,等.基于ARIMA和小波神经网络组合模型的交通流预测[J].计算机技术与发展,2017,27(1):169-172.
[8] SEO Y,KIM S,KISI O,et al.Daily water level forecasting using wavelet decomposition and artificial intelligence techniques[J].Journal of Hydrology,2015,520:224-243.
[9] 刘伟,李春青,张艳芬.基于小波分解和最小二乘支持向量机的COD预测[J].微电子学与计算机,2013,30(4):26-29.
[10] ZHANG Q,BENVENISTE A.Wavelet networks[J].IEEE Transactions on Neural Networks,1992,3(6):889-898.
[11] 杨颖,陈德华.基于小波神经网络的时间序列流数据的研究[J].计算机技术与发展,2006,16(6):193-195.
[12] 鞠琴,余钟波,郝振纯,等.小波网络模型及其在日流量预测中的应用[J].西安建筑科技大学学报:自然科学版,2009,41(1):47-52.
[13] 朱跃龙,李士进,范青松,等.基于小波神经网络的水文时间序列预测[J].山东大学学报:工学版,2011,41(4):119-124.
[14] 李代华,崔东文.相空间重构支持向量机在径流模拟中的应用研究[J].长江科学院院报,2013,30(10):21-26.
[15] 廖荣华,兰时勇,刘正熙.基于混沌时间序列局域法的短时交通流预测[J].计算机技术与发展,2015,25(1):1-5.
[16] FALAMARZI Y,PALIZDAN N,HUANG Y F,et al.Estimating evapotranspiration from temperature and wind speed data using artificial and wavelet neural networks(WNNs)[J].Agricultural Water Management,2014,140(3):26-36.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |