A multiple-kernel LSSVR method for separable nonlinear system identification-一种可分离非线性系统辨识的多核LSSVR方法.pdfVIP

JControl 20131 Theory Appl DOI10．1007／s1 1768—013-2035．9 A LSSVRmethodfor multiple·—kernel ’‘ ‘ ‘‘ ●1 ‘‘一 nonlinearldentilicatlon system Yanning WANG，Xuemei FAN CAIt，Hongqiao YE，Qinggang Xi’anResearchInstituteofHi·Tech．Xi’anShannxj710025．Chjna Abstraet：Insomenonlinear statevariablesfunction Callbe fromthe dynamicsystems，the usually separated control variables to function，whichmuchtroubletheidentificationofsuch wellsolvethis brings systems．To problem，animproved least vector with is andthemodelis tothe squaressupport regression(LSSVR)modelmultiple—kernelproposed applled nonlinear identification．Thismethodutilizestheexcellentnonlinear ofMorletwavelet separablesystem mappingability kernelfunctionand the andconffolvariables combinesstate informationintoakernel the wavelet matrix．Usingcomposite LSSVRincludestwononlinear variablesarethestatevariablesandthe ones kernel，the functions，whose