Scholarlink

<p>Coalbed methane (CBM) is a significant, widely distributed, and clean energy source, playing a crucial role in addressing the increasing global energy demand. Coal structure, a key parameter in the development of CBM, influences the main length and complexity of coal seams during the hydraulic fracturing process. Identifying coal structures through geophysical logging has become a hot topic due to the high cost and time-consuming nature of coal cores. This paper takes the No. 3 coal from the Zhengzhuang Block in the Qinshui Basin as a case study and proposes a coal structure identification method using acoustic time difference (AC), caliper log (CAL), natural gamma (GR), density (DEN), and deep lateral resistivity (LLD) as input features in a convolutional neural network (CNN) model. The study demonstrated that AC, CAL, and LLD increased with coal damage, while GR and DEN decreased. The trained CNN model achieved an accuracy of 97.8% on the training dataset and 92.3% on the testing dataset. Further validation through blind well tests on wells 64 and 65 yielded accuracies of 94.5% and 92.3%, respectively, demonstrating the CNN model’s strong generalization capability. The planar and sectional features of coal seams in the study area were analyzed: the undeformed coal was thicker in the southwest part of the study area, the thickness of cataclastic and granulated coal decreased gradually along the southwest direction, and the coal seams showed inhomogeneity in the vertical direction. This study provides an efficient and accurate approach for coal structure classification, offering a valuable tool for CBM exploration and development.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identification of Coal Structure Based on Geophysical Logging Data in Zhengzhuang Block, Southern Qinshui Basin, China: Investigation by Convolutional Neural Networks

  • 期刊论文

  • 发表日期：

    2025年6月6日

查看全文
  • Tingrui Wan,
  • Pengfei Ren,
  • Chuanbang Chen,
  • Hanghang Ping

摘要

Coalbed methane (CBM) is a significant, widely distributed, and clean energy source, playing a crucial role in addressing the increasing global energy demand. Coal structure, a key parameter in the development of CBM, influences the main length and complexity of coal seams during the hydraulic fracturing process. Identifying coal structures through geophysical logging has become a hot topic due to the high cost and time-consuming nature of coal cores. This paper takes the No. 3 coal from the Zhengzhuang Block in the Qinshui Basin as a case study and proposes a coal structure identification method using acoustic time difference (AC), caliper log (CAL), natural gamma (GR), density (DEN), and deep lateral resistivity (LLD) as input features in a convolutional neural network (CNN) model. The study demonstrated that AC, CAL, and LLD increased with coal damage, while GR and DEN decreased. The trained CNN model achieved an accuracy of 97.8% on the training dataset and 92.3% on the testing dataset. Further validation through blind well tests on wells 64 and 65 yielded accuracies of 94.5% and 92.3%, respectively, demonstrating the CNN model’s strong generalization capability. The planar and sectional features of coal seams in the study area were analyzed: the undeformed coal was thicker in the southwest part of the study area, the thickness of cataclastic and granulated coal decreased gradually along the southwest direction, and the coal seams showed inhomogeneity in the vertical direction. This study provides an efficient and accurate approach for coal structure classification, offering a valuable tool for CBM exploration and development.