%0 Journal Article %T 基于鄂尔多斯盆地西北缘地区烃源岩有机碳含量测井评价方法研究
Integrated Logging Assessment of Organic Carbon Content in Source Rocks: Northwestern Ordos Basin Applications %A 郭聪聪 %A 何之龙 %A 徐志峰 %A 王子璇 %A 宋姗 %J Journal of Oil and Gas Technology %P 211-222 %@ 2471-7207 %D 2025 %I Hans Publishing %R 10.12677/jogt.2025.472025 %X 烃源岩总有机碳(Total Organic Carbon, TOC)含量是非常规油气藏的重要参数之一，准确评价TOC含量对非常规油气藏的勘探开发具有重要意义。本文研究了两类TOC评价方法，一类方法基于物理模型评价烃源岩有机碳含量，包括∆LogR方法、变基线方法以及双差分方法(DD∆LogR)。相较于∆LogR方法，变基线方法改进了基线确定方法，提高了模型精度；双差分方法不仅考虑基线问题，还考虑到岩石矿物成分的变化和孔隙扩大的影响，在三种基于物理模型的TOC评价方法中效果最优。另一类方法基于机器学习评价烃源岩有机碳含量，本文对比了BP神经网络、LSTM神经网络、与物理模型结合的BP神经网络、与物理模型结合的LSTM神经网络，在四种神经网络方法中，与物理模型结合的LSTM神经网络评价TOC效果最优。本研究建立的综合评价方法体系，为鄂尔多斯盆地西北缘复杂地质条件下烃源岩有机质含量预测提供了创新性的技术路径。
The total organic carbon (TOC) content of source rocks is one of the important parameters for unconventional oil and gas reservoirs. Accurate evaluation of the TOC content is of great significance for the exploration and development of unconventional oil and gas reservoirs. This study investigates two types of TOC evaluation methods. For physical-model-based methods, it includes the ∆LogR method, the variable baseline method, and the double difference method (DD∆LogR). Compared with the ∆LogR method, the variable baseline method improves baseline determination and enhances model accuracy; the double-difference method (DD∆LogR) further considers rock mineral composition changes and pore expansion, outperforming the other two physical-model-based methods. Regarding machine-learning-based methods, a comparison of BP neural network, LSTM neural network, and their hybrid versions integrated with physical models shows that the LSTM neural network combined with physical models yields the most accurate TOC evaluation. The comprehensive evaluation framework established here offers an innovative solution for predicting source-rock organic matter content in the complex geological environment of the northwestern Ordos Basin, potentially improving resource assessment efficiency and accuracy in similar settings. %K TOC评价， %K ∆ %K LogR方法， %K 机器学习
TOC Evaluation %K ∆ %K LogR Method %K Machine Learning %U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=117217