首页|期刊导航|测井技术|形态加权表征的测井曲线深度校正方法

形态加权表征的测井曲线深度校正方法

FANG Yu XIAO Lizhi LUO Sihui LIU Jiaxiu LIAO Guangzhi ZHOU Jun ZHANG Juan

测井技术2025，Vol.49Issue(6)：836-844,856,10.

测井技术2025，Vol.49Issue(6)：836-844,856,10.DOI:10.16489/j.issn.1004-1338.2025.06.002

形态加权表征的测井曲线深度校正方法

Shape-Weighted Representation for Well-Log Depth Matching

FANG Yu ¹XIAO Lizhi ¹LUO Sihui ²LIU Jiaxiu ³LIAO Guangzhi ¹ZHOU Jun ⁴ZHANG Juan⁴

作者信息

1. State Key Laboratory of Petroleum Resources and Engineering,Beijing 102249,China||College of Geophysics,China University of Petroleum(Beijing),Beijing 102249,China
2. State Key Laboratory of Petroleum Resources and Engineering,Beijing 102249,China||College of Carbon Neutral Energy,China University of Petroleum(Beijing),Beijing 102249,China
3. State Key Laboratory of Petroleum Resources and Engineering,Beijing 102249,China||College of Artificial Intelligence,China University of Petroleum(Beijing),Beijing 102249,China
4. College of Artificial Intelligence,China University of Petroleum(Beijing),Beijing 102249,China||Logging Technology Research Institute,China National Logging Corporation,Xi'an,Shaanxi 710077,China||Well Logging Key Laboratory,China National Petroleum Corporation,Xi'an,Shaanxi 710077,China
折叠

摘要

Abstract

Depth matching of well logging curves serves as a critical foundation for geological modeling and accurate reservoir parameter evaluation.To address the limitations of existing methods in matching key geological features and capturing nonlinear variations,this study proposes a shape-weighted representation method for well-log curve depth matching.During training,a peak-valley weighted loss(PVWL)method is introduced to enhance the model's ability to fit critical geological features.Meanwhile,the multi-head attention mechanism of the Transformer effectively captures nonlinear patterns in logging sequences.In the inference stage,a feature-point matching module is incorporated to achieve high-precision depth alignment under morphological consistency constraints.The experimental dataset comprises conventional logging curves from ten field exploration wells in a block of the Ordos basin,using manually corrected curves by geological experts as ground truth.Results demonstrate that,while maintaining comparable global fitting performance(determination coefficient slightly decreases from 98.59%to 98.55%,and root-mean-square error increases marginally to 4.61),PVWL significantly improves local feature fitting accuracy(local root mean square error decreases from 2.88 to 2.38).Sequence length experiments reveal that a length of 3 yields optimal model performance.Compared with the recurrent neural network,gated recurrent unit,and long short-term memory network models,the proposed method exhibits clear advantages in both global fitting and alignment of key geological features.Ultimately,the corrected curves obtained via the feature-point matching module show high consistency with expert-corrected labels,validating the accuracy and robustness of the proposed approach.

关键词

人工智能/深度学习/深度校正/测井曲线/Transformer模型/损失函数/评价指标/解释评价/自然伽马曲线

Key words

artificial intelligence/deep learning/depth matching/well log/Transformer network/loss function/evaluation indicator/interpretation evaluation/natural gamma curve

分类

天文与地球科学

引用本文复制引用

FANG Yu,XIAO Lizhi,LUO Sihui,LIU Jiaxiu,LIAO Guangzhi,ZHOU Jun,ZHANG Juan..形态加权表征的测井曲线深度校正方法[J].测井技术,2025,49(6):836-844,856,10.

基金项目

国家重点研发计划课题"低场核磁共振宽频测量仪"(2023YFF0714102) （2023YFF0714102）

国家重点研发计划课题"复杂油气智能钻井理论与方法"(2019YFA0708301) （2019YFA0708301）

国家自然科学基金项目"钠氢双核-核磁共振弛豫机理与成像方法研究"(42474165) （42474165）

中国石油天然气集团公司-中国石油大学(北京)战略合作科技专项"物探、测井、钻完井人工智能理论与应用场景关键技术研究"(ZLZX2020-03) （北京）

测井技术

ISSN：1004-1338

访问量0

下载量0

段落导航