基于电成像测井的碳酸盐岩地层地质结构自动分类新方法  

Novel Automatic Classification Method for Geological Structures in Carbonate Formations Based on Electrical Imaging Logging

作　　者：付雅峰 黄科 朱涵斌 王慧 张旭 赵洁 肖霓 FU Yafeng;HUANG Ke;ZHU Hanbin;WANG Hui;ZHANG Xu;ZHAO Jie;XIAO Ni(Geological Research Institute,China National Logging Corporation,Xi'an,Shaanxi 710077,China;Well Logging Key Laboratory,China National Petroleum Corporation,Xi'an,Shaanxi 710077,China;Hangzhou Sumay Technology CO.LTD.,Hangzhou,Zhejiang 310007,China;Bomay Electric Industries CO.LTD.,Xi'an,Shaanxi 710076,China)

机构地区：[1]中国石油集团测井有限公司地质研究院,陕西西安710077 [2]中国石油天然气集团有限公司测井重点实验室,陕西西安710077 [3]杭州迅美科技有限公司,浙江杭州310007 [4]西安宝美电气工业有限公司,陕西西安710076

出　　处：《测井技术》2025年第1期68-76,共9页Well Logging Technology

基　　金：中国石油天然气集团有限公司科技项目“油基钻井液成像测井关键核心技术研究”(2024ZG45);中国石油集团测井有限公司科学研究与技术开发项目“测井油藏与地质研究”(CNLC2023-8B03)。

摘　　要：在碳酸盐岩地层的电成像测井资料处理中,泥岩层理、天然裂缝、诱导缝和孔洞等结构由于电阻率相近且相互重叠而难以区分,针对此问题,提出了一种基于骨架图和机器学习的自动分类方法。通过改进的K-means聚类算法从电成像数据中分割出感兴趣区域;利用图像细化和Hough变换获取骨架图的线段化描述,并基于空间几何关系实现线段分组;使用方向性区域生长算法分离重叠区域,提取独立连通区域的几何特征(如周长、面积、平均倾角等);基于这些特征训练极端梯度提升树模型,实现对不同地质结构的自动分类。通过碳酸盐岩的电成像测井数据验证,结果显示,该方法的分类精度与岩心观察结果的符合率大于80%。与现有研究相比,该文首次基于电成像图像实现了对碳酸盐岩地层泥岩层理、天然裂缝、诱导缝和孔洞等地质结构的自动分类,显著提高了分类效率和准确性。In the processing of electrical imaging logging data for carbonate formations,it is challenging to distinguish mudstone laminae,natural fractures,induced fractures,and vugs due to their similar resistivities and overlapping occurrences.To address this issue,this paper proposes an automatic classification method based on skeleton maps and machine learning.First,an improved K-means clustering algorithm is used to segment regions of interest from the electrical imaging data.Then,image thinning and the Hough transform are employed to obtain a line-segment representation of the skeleton map,and line segments are grouped based on spatial geometric relationships.Next,a directional region-growing algorithm is applied to separate overlapping regions and to extract geometric features(e.g.,perimeter,area,average dip,etc.)of each independent connected region.Finally,these features are used to train an extreme gradient boosting(XGBoost)model,enabling the automatic classification of different geological structures.Validation using actual carbonate logging data shows that the classification accuracy of this method exceeds an 80%agreement rate with core observations.Compared with existing research,this study is the first to achieve automatic classification of mudstone laminae,natural fractures,induced fractures,and vugs in carbonate formations based on electrical imaging logs,significantly improving classification efficiency and accuracy.

关 键 词：碳酸盐岩电成像 骨架图 方向性区域生长算法 极端梯度提升树 自动分类 

分 类 号：P631.84[天文地球—地质矿产勘探]