机器学习方法在浅层滩坝相薄储层孔隙度预测中的应用——以准噶尔盆地车排子地区白垩系为例  

Application of Machine Learning for Porosity Estimation of Beach and Bar Sand Bodies in a Lacustrine Basin:A case study of the Lower Cretaceous strata in Chepaizi area,Junggar Basin,NW China

作　　者：张宇航 时保宏[1,2] 张曰静 石好果[4] 文雯[5] 张杨[5] ZHANG YuHang;SHI BaoHong;ZHANG YueJing;SHI HaoGuo;WEN Wen;ZHANG Yang(School of Earth Sciences and Engineering,Xi'an Shiyou University,Xi'an 710065,China;Shaanxi Key Lab of Petroleum Accumulation Geology,Xi'an Shiyou University,Xian 710065,China;State Key Laboratory of Petroleum Resource and Prospecting,China University of Petroleum(Beijing),Beijing 102249,China;Research Institute of Exploration and Development,Shengli Oilfield Company,SINOPEC,Dongying,Shandong 257061,China;First Oil Prodction Plant of CNPC Qinghai Oilfield Company,Dunhuang,Gansu 736202,China)

机构地区：[1]西安石油大学地球科学与工程学院,西安710065 [2]西安石油大学陕西省油气成藏地质学重点实验室,西安710065 [3]中国石油大学(北京)油气资源与探测国家重点实验室,北京102249 [4]中国石化胜利油田分公司勘探开发研究院,山东东营257061 [5]中国石油青海油田分公司采油一厂,甘肃敦煌736202

出　　处：《沉积学报》2023年第5期1559-1567,共9页Acta Sedimentologica Sinica

基　　金：国家自然科学基金项目(41711530128);陕西省自然科学基金项目(2021JQ-587);油气资源与探测国家重点实验室开放课题基金(PRP/open-1609)。

摘　　要：准噶尔盆地车排子地区白垩系储层以滩坝相沉积为主,储层砂体薄,纵向变化快,孔隙度估算难度较大。基于Xgboost机器学习算法,根据取心井的岩心实测数据,结合其对应的测井数据,建立了测井孔隙度模型。结果表明,研究区对储层孔隙度影响较大的测井变量为自然伽马测井、声波测井、密度测井和冲洗带电阻率测井,其相关系数分别为0.38、0.42、0.28和0.32。基于特征测井数据,利用Xgboost算法预测的孔隙度与实测孔隙度吻合度较高,相关系数为0.92,均方差为0.20。此外,对近期钻探的新井储层孔隙度进行预测,结果表明孔隙度较高的井段与试油数据相吻合,从侧面反映了模型的可靠性。这一结果为研究区油气藏评价和后期油藏模型的建立提供基础数据,有利于提高研究区勘探的精度。同时,该模型也可用于类似滩坝相、砂体薄的沉积背景下储层孔隙度估算研究。The facies of Cretaceous reservoirs are beach and bar,and the sandstone reservoirs are characterized by thin and sharp vertical change in the Chepaizi area of the Junggar Basin.As a result,new challenges appear in esti⁃mating the porosity of reservoirs.In this study,based on the measured porosity in the laboratory and corresponding logging data,a new porosity estimation model was established using the extreme gradient boosting(Xgboost)ma⁃chine learning algorithm.The results show that the correlation coefficients between reservoir porosity and GR,AC,CNL,and RXO are 0.38,0.42,0.28,and 0.32,respectively,suggesting that porosity is influenced by the logging data in the study area.Based on the input logging data,the predicted porosity using the Xgboost algorithm matches the measured porosity,with a correlation coefficient of 0.92 and a mean squared error of 0.20.To test and verify the predicted results from the Xgboost mothed,we use the production test result as collateral evidence.The results show that the well sections with higher porosity match with the test data,indicating the reliability of the model.This result provides fundamental data for reservoir evaluation and modelling,improving the exploration accuracy in the study ar⁃ea.Furthermore,the model can be used in the study of reservoir porosity estimation in similar sedimentary environ⁃ments.

关 键 词：机器学习 孔隙度估算 滩坝相 白垩系 车排子凸起 

分 类 号：P613.18[天文地球—矿床学]