Scholarlink

<p>Conventional porosity inversion methods typically derive elastic parameters from pre-stack seismic inversion and then convert them to porosity through rock physics relationships. Due to inherent issues of such stepwise strategies—such as error accumulation, strong model dependence, and insufficient utilization of the correlational information in logging—this paper proposes a novel porosity inversion method based on joint dictionary learning and spatial structure constraints from post-stack seismic data. This method abandons the traditional stepwise inversion strategy and innovatively constructs a joint sparse representation framework. First, leveraging logging data from the target area, it employs joint dictionary learning technology to extract the features of acoustic impedance and porosity as well as the complex nonlinear mapping relationship between them, thereby forming a joint dictionary containing their intrinsic correlation as prior knowledge. Subsequently, this dictionary is incorporated into the post-stack seismic inversion process, enabling the simultaneous inversion of impedance and porosity, which fundamentally avoids the error propagation inherent in stepwise inversion. Furthermore, to improve the spatial continuity of the inversion results, the Gradient Structure Tensor (GST) technique is introduced to extract the spatial structural features of the subsurface media from seismic data as a constraint, effectively enhancing the horizontal continuity and geological plausibility of the porosity model. Experiments on the Marmousi theoretical model demonstrate that compared to traditional linear regression methods, the proposed method reduces the root mean square error (RMSE) of the inversion results by approximately 30%, and exhibits significantly superior spatial continuity compared to joint dictionary inversion results without spatial constraints. In a practical application to a fractured-vuggy carbonate reservoir in the Amu Darya Basin of Central Asia, this method successfully delineated richer and more detailed internal structures of pores and vugs, validating its effectiveness and application potential in solving practical, complex geological problems.</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

Porosity Inversion Using a Joint Dictionary and Spatial Structure with Simultaneous Constraints: An Application to Carbonate Fractured-Vuggy Reservoirs

  • 期刊论文

  • 发表日期：

    2026年4月23日

查看全文
  • Feng Ma,
  • Ming Lei,
  • Fei Xu,
  • Han-peng Cai,
  • Yu Liu,
  • Ting Cheng,
  • Yao-jun Wang,
  • Guang-min Hu

摘要

Conventional porosity inversion methods typically derive elastic parameters from pre-stack seismic inversion and then convert them to porosity through rock physics relationships. Due to inherent issues of such stepwise strategies—such as error accumulation, strong model dependence, and insufficient utilization of the correlational information in logging—this paper proposes a novel porosity inversion method based on joint dictionary learning and spatial structure constraints from post-stack seismic data. This method abandons the traditional stepwise inversion strategy and innovatively constructs a joint sparse representation framework. First, leveraging logging data from the target area, it employs joint dictionary learning technology to extract the features of acoustic impedance and porosity as well as the complex nonlinear mapping relationship between them, thereby forming a joint dictionary containing their intrinsic correlation as prior knowledge. Subsequently, this dictionary is incorporated into the post-stack seismic inversion process, enabling the simultaneous inversion of impedance and porosity, which fundamentally avoids the error propagation inherent in stepwise inversion. Furthermore, to improve the spatial continuity of the inversion results, the Gradient Structure Tensor (GST) technique is introduced to extract the spatial structural features of the subsurface media from seismic data as a constraint, effectively enhancing the horizontal continuity and geological plausibility of the porosity model. Experiments on the Marmousi theoretical model demonstrate that compared to traditional linear regression methods, the proposed method reduces the root mean square error (RMSE) of the inversion results by approximately 30%, and exhibits significantly superior spatial continuity compared to joint dictionary inversion results without spatial constraints. In a practical application to a fractured-vuggy carbonate reservoir in the Amu Darya Basin of Central Asia, this method successfully delineated richer and more detailed internal structures of pores and vugs, validating its effectiveness and application potential in solving practical, complex geological problems.