基于神经网络聚类分析的深层页岩储层岩相识别——以川南筇竹寺组为例

Lithofacies Identification in Deep Shale Reservoirs Via Neural Network Clustering Analysis: A Case Study of the Qiongzhusi Formation in the Southern Sichuan Basin

董晓霞 冯少柯

DONGXiaoxia FENGShaoke

中国石化西南油气分公司, 四川 成都 610041 成都理工大学能源学院, 四川 成都 610059

Southwest Petroleum Branch, SINOPEC, Chengdu, Sichuan 610041, China College of Energy, Chengdu University of Technology, Chengdu, Sichuan 610059, China

随着中国石化西南油气分公司下寒武统筇竹寺组页岩气勘探的重大突破，四川海相页岩气勘探热点逐渐从龙马溪组向筇竹寺组进行转变。因此，如何对页岩岩相进行准确识别是现阶段勘探工作中尚待解决的难题。为了解决这一问题，根据岩芯样品的有机质含量和X射线衍射实验结果，将筇竹寺组深层页岩储层划分为5种岩相(富有机质的粉砂质页岩和含钙粉砂质页岩、贫有机质的粉砂质页岩、含钙粉砂质页岩和黏土质页岩)。在三角图岩相划分和岩相特征分析的基础上，基于神经网络聚类分析理论建立了深层页岩气储层岩相识别工作流和模型，测试、验证和训练数据集的混淆矩阵结果均大于88%，识别准确性高。利用其模型对Z2井岩相进行了识别，比传统的岩相方法更加准确、高效，有助于研究区深层页岩气储层的高效开发，也为深层—超深层页岩气储层岩相识别研究提供了新思路。

With the significant breakthrough in shale gas exploration of the Lower Cambrian Qiongzhusi Formation by SINOPEC Southwest Oil and Gas Company, the hot spot of marine shale gas exploration in Sichuan is gradually shifting from the Longmaxi Formation to the Qiongzhusi Formation. Therefore, how to accurately identify shale lithology is a difficult problem that needs to be solved in current exploration work. To address this issue, based on the Total organic matter content and XRD experimental results of core samples, the deep shale reservoirs of the Qiongzhusi Formation were divided into five lithofacies (organic rich silty shale, organic rich calcium containing silty shale, organic poor silty shale, organic poor calcium containing silty shale, and organic poor clayey shale). Based on the triangulation of lithofacies and analysis of lithofacies characteristics, a workflow and model for identifying lithofacies in deep shale gas reservoirs were established using neural network clustering analysis theory. The confusion matrix results of the testing, validation, and training datasets were all greater than 88%, indicating high recognition accuracy. The identification of lithology in Well Z2 using its model is more accurate and efficient than traditional lithology methods, which is helpful for the efficient development of deep shale gas reservoirs in the study area and provides new ideas for lithology identification research in deep ultra deep shale gas reservoirs.

10.11885/j.issn.1674 5086.2024.10.31.01