In this paper, taking the shale of Chang 7-Chang 9 oil formation in Yanchang Formation in the southeastern Ordos Basin as an example, through the study of shale heterogeneity characteristics, starting from the preprocessing of supervision data set, a logging interpretation method of total organic carbon content (TOC) on the lithofacies-based Categorical regression model (LBCRM) is proposed. It is show that: 1) Based on core observation, and Differences of sedimentation and structure, five lithofacies developed in the Yanchang Formation: shale shale facies, siltstone/ultrafine sandstone facies, tuff facies, argillaceous shale facies with silty lamina and argillaceous shale facies with tuff lamina. 2) The strong heterogeneity of shale makes it difficult to accurately explain the TOC distribution of shale intervals in the application of model-based interpretation methods. The LBCRM interpretation method based on the understanding of shale heterogeneity can effectively reduce the influence of formation factors other than TOC on the prediction accuracy by studying the characteristics of shale heterogeneity and constructing a TOC interpretation model for each lithofacies category. At the same time, the degree of unbalanced distribution of data is reduced, so that the data mining algorithm achieves better prediction effect. 3) The interpretability of lithofacies logging ensures the wellsite application based on the classification and regression model of lithofacies. Compared with the traditional homogeneous regression model, the prediction performance has been greatly improved, TOC segment prediction is more accurate. 4) The LBCRM method based on shale heterogeneity can better understand the reasons for the deviation of the traditional model-based interpretation method. After being combined with the latter, it can make logging data provide more useful information.
Full Text Finder