Font Size: a A A

Study On The Evolution Of Runoff And Prediction Method Of Huangshui River Basin

Posted on:2019-07-05Degree:MasterType:Thesis
Country:ChinaCandidate:C SuFull Text:PDF
GTID:2310330569997649Subject:Structure engineering
Abstract/Summary:PDF Full Text Request
For the past few years,with the increase of global climate change and human activities,the runoff of Huangshui River has been decreasing year by year.In the face of such situation,we must ahead of various analysis forecast and control the runoff changes.This paper takes the measured annual runoff data of the four main hydrological sites of Huangshui River(Shiyazhuang station,Xining station,Ledu station,Minhe station)as the research object.Application of traditional methods(cumulative departure method,wavelet analysis,Mann-Kendall test,T test)and time-frequency analysis method(empirical mode decomposition,EMD)to analyze the inter annual variation characteristics of the year,and the characteristics of wetness-dryness,trend,cycle and mutation.In the runoff prediction section,three methods of support vector machine(SVM),EMD+SVM and improved EMD+SVM were used to predict the results,and three methods were used to predict the effect by relative error size.The dry year of Huangshui basin is more and more long time,the use of traditional methods,5?15?30 years that flow exists in the cycle of change,runoff presents a downward trend,runoff decline trend let up after 2000;the time-frequency method(EMD)was used to obtain the periodic variation of the runoff in 3.08,7.3,11.5 and 33 year,which is closer to the sunspot cycle(11.5 years)and el nino phenomenon(2 to 7 years).In the three prediction methods,the relative error of SVM alone is about 13.5%,and the accuracy of the prediction point the minimum is 4.68%,the maximum is 22.60%.The relative error of EMD+SVM is about 8.5%,and the accuracy of the prediction point the minimum is 3.59%,the maximum is 11.57%.The relative error of improved EMD+SVM is about 6.2%,and the accuracy of the prediction point the minimum is 0.09%,and the maximum is 10.27%.It can be found that improved EMD+SVM prediction is better.In this paper,using the time-frequency analysis method to runoff series processed before further analysis,and put forward to improve the EMD for the runoff data of smooth more ideal,and making cycle,trend,improve the precision of the predicted,ect.But the article also has many disadvantages.We should continue to look for more effective and reliable medium-and-long-term forecast methods,and do well in water resources management and scientific planning.
Keywords/Search Tags:Huangshui basin, EMD, Modal aliasing, Runoff forecasting method
PDF Full Text Request
Related items