20999.基于測(cè)井?dāng)?shù)據(jù)的巖性識(shí)別方法研究

1、論文題目：基于測(cè)井?dāng)?shù)據(jù)的巖性識(shí)別方法研究專業(yè)：應(yīng)用數(shù)學(xué)碩士生：李保霖(簽名)指導(dǎo)教師：龍熙華(簽名)摘要隨著測(cè)井技術(shù)的發(fā)展，綜合應(yīng)用各種測(cè)井?dāng)?shù)據(jù)，成為更快速，更準(zhǔn)確的獲得巖性信息最主要的途徑。巖性識(shí)別是測(cè)井?dāng)?shù)據(jù)解釋中最關(guān)鍵的一環(huán)。傳統(tǒng)的識(shí)別方法效率慢、精度低、人為因素大、不利于實(shí)際工程的應(yīng)用。因此，構(gòu)造一種速度快，識(shí)別率高，泛化能力好的巖性自動(dòng)識(shí)別法成為測(cè)井?dāng)?shù)據(jù)解釋領(lǐng)域迫切的需求。圍繞巖性識(shí)別的目的，本文主要完成三方面內(nèi)容：測(cè)井?dāng)?shù)據(jù)預(yù)處

2、理，測(cè)井相分析，巖性識(shí)別方法研究。測(cè)井?dāng)?shù)據(jù)預(yù)處理是在測(cè)井?dāng)?shù)據(jù)解釋之前，為消除非地層因素的影響而必要完成的基礎(chǔ)性工作；測(cè)井相分析是建立在多源信息融合的測(cè)井曲線自動(dòng)分層基礎(chǔ)上，為構(gòu)建全區(qū)域測(cè)井相巖性數(shù)據(jù)庫(kù)而必須執(zhí)行的過程，它為巖性識(shí)別提供樣本選取的標(biāo)準(zhǔn)和依據(jù)；巖性識(shí)別方法研究針對(duì)三種神經(jīng)網(wǎng)絡(luò)：遺傳BP網(wǎng)絡(luò)，F(xiàn)CM_PNN網(wǎng)絡(luò)，遺傳優(yōu)化RBPNN網(wǎng)絡(luò)，進(jìn)行了算法構(gòu)造和巖性識(shí)別應(yīng)用對(duì)比。結(jié)果表明：三種網(wǎng)絡(luò)對(duì)巖性識(shí)別的應(yīng)用均是可行的，但遺傳BP速

3、度較慢，F(xiàn)CM_PNN識(shí)別正確率相對(duì)較低，都不能完全達(dá)到實(shí)際工程的應(yīng)用規(guī)范。為此，構(gòu)造全結(jié)構(gòu)遺傳優(yōu)化的RBPNN網(wǎng)絡(luò)模型。它通過采用遺傳算法搜索使得RBPNN訓(xùn)練法誤差最小的最優(yōu)隱中心矢量，以及相匹配的核函數(shù)控制參數(shù)，優(yōu)化網(wǎng)絡(luò)結(jié)構(gòu)，增加網(wǎng)絡(luò)穩(wěn)定性。相對(duì)遺傳BP和FCM_PNN，遺傳優(yōu)化RBPNN在保證收斂速度和識(shí)別精度的同時(shí)能夠?qū)崿F(xiàn)巖性信息更快速和更準(zhǔn)確的識(shí)別，且網(wǎng)絡(luò)結(jié)構(gòu)簡(jiǎn)單，推廣泛化能力強(qiáng)，能為實(shí)際測(cè)井?dāng)?shù)據(jù)解釋領(lǐng)域所應(yīng)用?；跍y(cè)井?dāng)?shù)據(jù)

4、的巖性識(shí)別方法研究為測(cè)井?dāng)?shù)據(jù)解釋提供了科學(xué)的方法支持和理論依靠，對(duì)整個(gè)油田地質(zhì)勘探領(lǐng)域具有重要的實(shí)際意義。關(guān)鍵詞：巖性識(shí)別；測(cè)井?dāng)?shù)據(jù)；BP神經(jīng)網(wǎng)絡(luò)；FCM_PNN神經(jīng)網(wǎng)絡(luò)；RBPNN神經(jīng)網(wǎng)絡(luò)研究類型：應(yīng)用研究Subject:TheMethodsResearchofLithologyIdentificationBasedonLoggingDataSpecialty:ApplicationMathematicsName:LIBaolin(S

5、ignature)Instruct:LONGXihua(Signature)ABSTRACTWiththedevelopmentoftheloggingtechnologythecomprehensiveapplicationofallstsofloggingdatahasbecamethemainwayfgettinginfmationoflithologyrapidlyaccurately.Lithologyidentificati

6、onisthemostkeyringofloggingdatainterpretation.Thetraditionalmethodshaveslowefficiencylowaccuracyhumanfactssothatnotconducivetotheactualengineeringapplication.Therefeconstructingaspeedhighrecognitionratestronggeneralizati

7、onabilitylithologyidentificationmethodhasbeenthepressingneedsininterpretationfieldofloggingdatas.Aroundthepurposeoflithologyidentificationthispapermainlycompletesthreeareas:loggingdatapretreatmentlogfaciesanalysisthemeth

8、odsresearchoflithologyidentification.loggingdatapretreatmentisabasicwkbefecompletingofloggingdatainterpretationisfeliminatingtheinfluencesofunfmationfacts.Logfaciesanalysisbasedonloggingcurveslayeringofthemultipleinfmati

9、onfusionisanecessaryprocessfconstructinglogfacieslithologydatabaseinthewholeregional.ThemethodsresearchoflithologyidentificationcompletealgithmconstructionapplicationcontrastofthreeneuralwkswhichcontainGABPneuralwkFCM_PN

10、NneuralwkgeicoptimizedRBPNNneuralwk.Theresultsshow:theapplicationsofthreemethodsinlithologyidentificationareaarefeasiblebutthereexitslowspeedtoGABPrelativelylowrecognitionaccuracytoFCM_PNN.Thustheycannotachievethestardof

11、practicalengineeringapplication.SostructuringgeicoptimizedRBPNNwkmodel.ItissearchedftheoptimizedhiddencentervectsmatchingkernelfunctioncontrolparametersoftheRBPNNbyusingthegeicalgithmwhichmustbesatisfiedminimumerrofRBPNN

12、training.RelativetotheBPwkFCM_PNNwkThismethodcannotnolyguaranteeconvergencespeedtheaccuracybutalsocanidentifylithologyinfmationsmequicklymeaccuratly.ithassimplewkstructurestrongpromotiongeneralizationability.Thatmakeitca