高分辨岩心曲线构建在测井岩性识别中的应用  

Application of High-Resolution Core Curve Construction in Logging Lithology Identification

作　　者：由婷 朱天怡 徐鹏晔 王树华 贺兴[2] 于立军[4] YOU Ting;ZHU Tianyi;XU Pengye;WANG Shuhua;HE Xing;YU Lijun(Research Institute of Carbon Neutrality,Shanghai Jiao Tong University,Shanghai 200240,China;School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Research Institute of Exploration&Development,Shengli Oilfield Company,SINOPEC,Dongying,Shandong 257099,China;College of Smart Energy,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学碳中和发展研究院,上海200240 [2]上海交通大学电子信息工程学院,上海200240 [3]中国石油化工股份有限公司胜利油田分公司勘探开发研究院,山东东营257099 [4]上海交通大学智慧能源创新学院,上海200240

出　　处：《测井技术》2023年第2期138-145,共8页Well Logging Technology

基　　金：中国石油化工股份有限公司“老油田开发大数据应用系统集成与示范应用”(P20071-4)。

摘　　要：在利用测井资料进行岩性识别时,为提高岩性识别的准确率通常会结合钻井取心实验,但取心过程复杂、成本高,油田的取心井数据一般较少,而人工智能、特别是机器学习技术的发展和应用为测井岩性识别提供了新的技术途径。以民丰洼陷北带盐22井区为例,采用机器学习算法开展地层岩性的自动识别,通过学习低分辨率(深度间隔0.125 m)的测井曲线数据,拟合出需要在实验室测试所得的高分辨率(深度间隔0.01 m)岩心伽马曲线,并将该曲线作为特征之一输入不同模型中进行地层岩性识别。结果显示,在多种机器学习模型中,拟合的高分辨率岩心伽马曲线不仅可以提高岩性识别的精确度,还可替代实测岩心伽马曲线在岩性识别中应用。其中,XG Boost模型表现最为突出,其岩性识别精确度最高为94.39%,为测井岩性识别提供了基于机器学习算法的有益探索。Stratum lithology identification based on logging curves is an important direction in the interpretation of logging data,it is usually combined with drilling and coring experiments.However,due to the complexity and high cost of the coring process,there is generally less data available from coring wells in oilfields.The development and application of artificial intelligence,especially machine learning technology,has provided new technical ways to solve the logging lithology identification problem.The research in this paper takes Yan22 well area in the northern belt of Minfeng sag as an example.We fit a high-resolution(depth interval of 0.01 m)core gamma curve,which usually needs to be obtained from laboratory tests,by learning low-resolution(depth interval of 0.125 m)logging curve data.Input the curve as one of the features into different models for lithology identification in the study of logging lithology identification based on machine learning algorithms.The results show that the fitted high-resolution core gamma curve can not only improve the accuracy of lithology identification as one of the new features,but also replace the measured core gamma curve features with maintaining a slightly higher or similar lithology identification accuracy in multiple machine learning models.In particular,the XG Boost model has the most outstanding performance,and its lithology recognition accuracy can be increased up to 94.39%,which provides a new and useful exploration for the lithology recognition of logging based on machine learning algorithms.

关 键 词：测井评价 岩性识别 机器学习 岩心曲线 XG Boost 民丰洼陷 

分 类 号：P631.84[天文地球—地质矿产勘探]