資源描述:
《基于小波理論的短時(shí)交通流預(yù)測(cè)方法研究》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、長(zhǎng)沙理工大學(xué)碩士學(xué)位論文基于小波理論的短時(shí)交通流預(yù)測(cè)方法研究姓名:喻丹申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):交通運(yùn)輸規(guī)劃與管理指導(dǎo)教師:吳義虎20080320摘要實(shí)時(shí)、準(zhǔn)確的短時(shí)交通流預(yù)測(cè)是智能交通控制與管理的基礎(chǔ),許多預(yù)測(cè)方法被提出,但是,因未考慮短時(shí)交通流中不確定干擾因素的影響,或者將各干擾簡(jiǎn)單復(fù)合統(tǒng)一處理,預(yù)測(cè)結(jié)果準(zhǔn)確性較差。本文從短時(shí)交通流特性分析出發(fā),利用PCA主成分分析法和分形理論驗(yàn)證短時(shí)交通流是一組包含干擾信號(hào)的混沌序列,具有最小可預(yù)測(cè)周期。短時(shí)交通流成分復(fù)雜,不同特性信號(hào)成分在預(yù)測(cè)中的作用不同,因此本文基于改進(jìn)的Mallat算法進(jìn)行小波分解和單支重構(gòu),將短時(shí)交通流分離成低頻
2、確定信號(hào)、高頻混沌信號(hào)和高頻干擾信號(hào)。對(duì)各分解信號(hào),構(gòu)造雙層小波網(wǎng)絡(luò)分別預(yù)測(cè):第一層小波網(wǎng)絡(luò)WNN.1用于低頻確定信號(hào)和高頻干擾信號(hào)的預(yù)測(cè);第二層小波網(wǎng)絡(luò)WNN.2用于高頻混沌信號(hào)的預(yù)測(cè)。最后,將各分解信號(hào)預(yù)測(cè)值迭加以獲得包含原始信號(hào)所有特性成分的預(yù)測(cè)值。算例研究表明,本文提出的雙層小波網(wǎng)絡(luò)短時(shí)交通流預(yù)測(cè)法具有較高預(yù)測(cè)精度和較快的預(yù)測(cè)速度。關(guān)鍵詞:短時(shí)交通流;雙層小波網(wǎng)絡(luò);多分辨;信號(hào)分解;混沌AbstractThecontrollingandmanagementofintectlencetrafficisbasedonthereal-timeandaccurateforec
3、astingofshort·termtrafficflow.Agreatmanyofforecastingapproachesforshort—termtrafficflowareadvanced.Buttheprecisionoftheseforecastingapproachesaredissatisfiedduetotheinfluenceofalltheinterferencedon’tbeconsideredfullyorthecombinedactionofinterferenceistransactedbysinglemethod.Sothecharacteri
4、sticofshort—termtrafficflowisanalyzedfirstbymethodnamedPrincipalComponentAnalysisandFractalTheoryinthepaper.Wecouldconcludethattheshort—termtrafficflowpossessedminimalpredictablecycleisasetofchaotictimeserialcontainingmuchinterference,andthedifferentcomponentinshort·termtrafficflowowneddiff
5、erentcharacteristichasdifferentinfluencefortheforecastingresults.SotheimprovedMallatalgorithmisadvancedinthepapertocompleteinformationseparation,whichisamethodusedforshort—termtrafficflowwaveletdecompositionandsingle-branchreconstruction,technically.Andtheshort-termtrafficflowtimeserialrese
6、archedinthepapercouldbeseparatedintolow·ffequencydetermineinformation,high·frequencychaoticinformationandhigh-frequencyinterferenceinformation.Thenadouble·layerwaveletnetworkisestablishedtoforecasttheinformationseparated:thefirstlayerofwaveletnetworkismarkedasWNN·1,whichisusedtoforecastthel
7、ow-·frequencydetermineinformationandhigh·-frequencyinterferenceinformation;thesecondlayerofwaveletnetworkiSmarkedasWNN·2,whichisusedtoforecastthehigh-矗equencychaoticinformation.Intheend,wecouldgettheforecastingresultsofshort—termtrafficflowbysuperposingt