Scholarlink

<p>To address the characteristic of frequent lithological alternations in the continental shale of the Songliao Basin in China and meet the refined requirements of reservoir modeling, it is necessary to establish a higher-precision lithology identification method. This study conducted scratch tests on shale reservoir cores from the 2360m–2409m interval of the Qingshankou Formation in the Songliao Basin, Jilin, obtaining nine mechanical characteristic parameters, including hardness, compressive strength, and Poisson’s ratio. By integrating convolutional neural network (CNN) and auto-encode network (AE), a novel lithology identification method based on scratch data was proposed. The optimal lithology identification scale was selected, and the performance of this method was compared with that of other neural network approaches. The results demonstrate that when the identification scale is set at 20 × 9, the test dataset achieves an accuracy of 89.58%, with recall rates exceeding 84% across all lithology recognitions, outperforming other identification scales. The convolutional autoencoder network (CAE) exhibits superior accuracy and recall rates in lithology identification compared to other neural networks, enabling a more precise representation of the actual lithological characteristics. This study provides a novel methodological approach for reservoir lithology identification and lays a foundation for modeling fracture propagation in heterogeneous shale reservoirs.</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

Convolutional autoencoder network lithology recognition based on scratch tests

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月19日

查看全文
  • Suling Wang,
  • Zhihui Ren,
  • Kangxing Dong,
  • Yanchun Li,
  • Jinbo Li,
  • Pengyun Wen,
  • Ruyi Qu,
  • Tingting Li,
  • Zhennan Wen

摘要

To address the characteristic of frequent lithological alternations in the continental shale of the Songliao Basin in China and meet the refined requirements of reservoir modeling, it is necessary to establish a higher-precision lithology identification method. This study conducted scratch tests on shale reservoir cores from the 2360m–2409m interval of the Qingshankou Formation in the Songliao Basin, Jilin, obtaining nine mechanical characteristic parameters, including hardness, compressive strength, and Poisson’s ratio. By integrating convolutional neural network (CNN) and auto-encode network (AE), a novel lithology identification method based on scratch data was proposed. The optimal lithology identification scale was selected, and the performance of this method was compared with that of other neural network approaches. The results demonstrate that when the identification scale is set at 20 × 9, the test dataset achieves an accuracy of 89.58%, with recall rates exceeding 84% across all lithology recognitions, outperforming other identification scales. The convolutional autoencoder network (CAE) exhibits superior accuracy and recall rates in lithology identification compared to other neural networks, enabling a more precise representation of the actual lithological characteristics. This study provides a novel methodological approach for reservoir lithology identification and lays a foundation for modeling fracture propagation in heterogeneous shale reservoirs.