鄂尔多斯盆地保德区块二叠系太原组—山西组主采煤层脆性评价——基于卷积神经网络方法

Brittleness evaluation of main coal seams in Permian Taiyuan-Shanxi formations, Baode block, Ordos Basin: based on a convolutional neural network method

张庆丰 李子玲 张继坤 郝帅 孙晓光 尚延洁 左运

ZHANG Qingfeng LI Ziling ZHANG Jikun HAO Shuai SUN Xiaoguang SHANG Yanjie ZUO Yun

中石油煤层气有限责任公司, 北京 100028

PetroChina Coalbed Methane Co., Ltd., Beijing 100028, China

鄂尔多斯盆地东北缘保德区块二叠系太原组—山西组煤层具有丰富的煤层气资源，但井间产能差异大，其重要原因在于储层脆性区域差异导致的强非均质性特征。岩石力学参数法是常用的储层脆性评价方法。研究岩石力学参数和脆性可为压裂改造提供重要参数基础，但当前方法多借助经验公式，评价精度有限。基于卷积神经网络方法，构建实验获取的弹性模量、泊松比与多测井曲线转换模型，建立了岩石力学剖面，进而实现脆性定量评价。结果显示：基于卷积神经网络预测含煤层系岩石力学参数适用性较好。保德区块主采4+5#和8+9#煤层脆性指数均整体较低，4+5#煤层较8+9#煤层脆性指数值略高，两套主采煤层平面分布具有一定的相似性，均在研究区中部和东南部脆性值低。矿物成分差异影响岩石脆性，石英含量越高，弹性模量和脆性指数越大，具有线性关系。

The coal seams of the Permian Taiyuan-Shanxi formations in the Baode block of the northeastern margin of the Ordos Basin have abundant coalbed methane resources. However, the productivity varies greatly among wells, mainly attributed to the strong heterogeneity caused by regional differences in reservoir brittleness. Rock mechanical parameter method is commonly used to evaluate reservoir brittleness. Studying rock mechanical parameters and brittleness can provide an important basis for fracturing modification. However, current methods mostly rely on empirical formulas, leading to limited evaluation accuracy. In this study, a convolutional neural network (CNN) was utilized to construct a conversion model between experimentally obtained elastic modulus, Poisson’s ratio, and multi-logging curves. Based on this method, rock mechanical profiles were further established, enabling quantitative evaluation of brittleness. The results indicated that CNN-based predictions of rock mechanical parameters had good applicability for coal-bearing layers. The brittleness indices the main coal seams, 4+5# and 8+9#, in the Baode block were generally low. The brittleness index of the 4+5# seam was slightly higher than that of the 8+9# seam. Both seams exhibited similar spatial distributions, with low brittleness values in the central and southeastern parts of the study area.Differences in mineral composition affected rock brittleness. Higher quartz content was linearly correlated with greater elastic modulus and brittleness index.

https://doi.org/10.11781/sysydz2025010204