基于BO-XGBoost优化方法的砂岩厚度预测方法研究与应用  

Research and application of sandstone thickness prediction method based on BO-XGBoost optimization method

作　　者：刘烜良 张军华[1] 白青林 王福金 刘中伟 焦红岩 LIU Xuanliang;ZHANG Junhua;BAI Qinglin;WANG Fujin;LIU Zhongwei;JIAO Hongyan(School of Geosciences,China University of Petroleum(East China),Qingdao 266580,China;Xianhe oil production plant,Shengli Oilfield Company,SINOPEC,Dongying,257068,China)

机构地区：[1]中国石油大学(华东)地球科学与技术学院,青岛266580 [2]中国石化胜利油田分公司现河采油厂,东营257068

出　　处：《物探化探计算技术》2024年第2期146-153,共8页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：国家自然科学基金项目(42072169);胜利油田项目(30200003-21-ZC0631)。

摘　　要：河道砂体预测在油气勘探中具有十分重要的作用,但在实际勘探开发中,复杂河道砂体地层埋深及河道本身结构的复杂性会使其井震关系较差,进而导致砂岩预测精度较低。针对该问题,笔者利用地震多属性信息,发挥密井网优势,利用BO-XGBoost砂岩厚度预测方法,在验证集占比为25%时取得最佳的预测效果,且预测效果要好于常规SVM和XGBoost方法。研究方法对同类复杂储层的砂岩厚度预测,有借鉴意义。Channel sand body prediction plays a vital role in oil and gas exploration,but in actual exploration and development,the buried depth of complex channel sand body and the complexity of channel structure will make its well seismic relationship poor,leading to low sandstone prediction accuracy.Given this problem,this paper uses the seismic multi-attribute information,gives full play to the advantages of the dense well pattern,and uses the BO-XGBoost sandstone thickness prediction method to achieve the best prediction effect when the validation set proportion is 25%.The prediction effect is better than the conventional SVM and XGBoost methods.The research method can be used for reference in sandstone thickness prediction of similar complex reservoirs.

关 键 词：贝叶斯优化 XGBoost 砂体 厚度预测 

分 类 号：P631.4[天文地球—地质矿产勘探]