基于FCMFS特征選擇算法的煤層氣壓裂效果預(yù)測(cè)

打開文本圖片集

CoalbedMethaneFracturingEffectPredictionBased on FCMFS Feature Selection Algorithm

MIN Chao1，2，3*， GUO Xing1，2，HUA Qing4， ZHANG Na4 ， ZHANG Xinhui1， 2

1.School of Science，Southwest Petroleum University，Chengdu， Sichuan 61o5oo，China 2.Institute forArtificial Intelligence，Southwest Petroleum University，Chengdu，Sichuan 61o5oo，China 3.StateKeybatoOdeoloEploatiousttroleesitygdu 4.ChongqingGas Mine，SouthwestOiland GasfieldCompany，PetroChina，Jiangbei，Chongqing 40o7oo，China

Abstract：Itisdificulttoanalyzethenonlinearrelationshipbetweenthefracturingeffectandcharacteristicsofcoalbedmethane fromthe mechanism level.Aimingat theproblem，the internalrelationship between thecharacteristicsofcoalbed methane fracturing effectisstudied，andaprediction methodofcoalbedmethane fracturingefectbasedonFCMFSfeatureselection algorithm is proposed.The methodusesfuzzycomprehensive evaluation tocalibrate thelabel，anduses genetic programming andXGBoostalgorithmtoconstructandscreenthecharacteristicsofinfluencingfactors，including twonewstructuralfeatures （stressratioandenticfactorsofgeologicalconstruction）ndsixharacterstcsofperforatiosectiontickne，peability fracturepressure，coalstructure，gassaturationandsandstrength.Theexperimentalresultsshowthatbasedontheeightfeatures constructedandscreenedbytheFCMFSfeatureselectionalgorithm，combinedwithavarietyofmachinelearningalgorithms to predictthefectofoalbedmethanefracturing，theacuracyrcallateandFlassificationevaluationindicatorsareiproed by about 5%～10% .Among them，the Deep Forest algorithm has the best prediction clasification effect on the training set and the test set，and the three classification evaluation indicators are all above 95% and 80%

Keywords：coalbed methane;fracturing performance; main controlling factor;gene programming; DeepForest model 網(wǎng)絡(luò)出版地址：http：//link.cnki.net/urlid/51.1718.TE.20240927.1400.016

引言

煤層氣作為一種儲(chǔ)量豐富的非常規(guī)天然氣資源，在中國能源結(jié)構(gòu)的優(yōu)化調(diào)整中發(fā)揮著重要作用[1-2]。（剩余12965字）