基于改进SMOTE和随机森林算法的致密砂岩成岩相测井解释方法

打开文本图片集

A Logging Interpretation Method for Tight Sandstone Diagenetic Facies Based on Improved SMOTE and Random Forest Algorithm

ZHEN Yan1，2， KANG Jintao1，2， ZHAO Xiaoming1，2*，GE Jiawang1，2，DAI Maolin1，2

1.SchoolofGeoscience andTechnology，SouthwestPetroleum University，Chengdu，Sichuan6o5o，Cina 2.SichuanKeyLaboratoryofNaturalGasGeology，Chengdu，Sichuan6lo5oo，China

Abstract：Diageneticlogginginterpretationisthekeytopredicthigh-qualitytightsandstonereservoirs.Comparedwithconventionalmathematicalstatisticalmethods，machinelearning methodscaneffectivelyimprovetheaccuracyofdiageneticfacies logginginterpretation.However，duetotheinsuffcientnumberofsamples，therearestilsome multi-solutionsintheinterpretationresults.Inorder toeffectivelysolve theproblemofsampledataimbalanceinthelogging interpretationofdiagenetic facies，thispaperconsidersthespatialconstraintsof new samplesonthebasisoftheclasical SMOTE（Synthetic Minority Over-samplingTechnique）algorithm.This paper puts forwarda RESMOTE（Repeat SMOTE）algorithm，which adds the few classamples intheimbalanceddataandusestherandomforestmodeltoidentifyandexplainthediageneticfacies.TheexperimentalresultsshowthattheRESMOTEalgorithmisbetter thanclassicalSMOTEalgorithm，Borderlie-SMOTEalgorithmand ADASYNalgorithm，and theaccuracyofrandom forest model is improved from 77.27% to 91.06% .TheRESMOTEalgorithm canensuretheaccuracyof newdata，efectively solve the problemofover-fting and lowaccuracyin conventional logging lithofacies identificationandclassfication methods，andhasimportantappicationvalueforthepredictionofhighqualitytight sandstone reservoirs.

Keywords： tight sandstone; diagenetic facies;RESMOTE algorithm; random forest; log interpretatio 网络出版地址：http：//link.cnki.net/urlid/51.1718.TE.20250701.1607.002

引言

中国致密砂岩油气资源分布广泛，总量丰富，在四川、塔里木、松辽、鄂尔多斯、渤海湾和东海等地都有巨大的开发潜力，中国也成为除美国和加拿大之后第三大致密砂岩气生产大国[1]。（剩余17706字）