Automatic Lithology Identification Based on Machine Learning: A Case Study of TZ4 Well（基于机器学习的岩性自动识别： TZ4 井案例研究）
发布时间：2024-01-08
元数据
名 称
Automatic Lithology Identification Based on Machine Learning: A Case Study of TZ4 Well（基于机器学习的岩性自动识别： TZ4 井案例研究）
科技资源标识
CSTR:11738.14.NCDC.XDA14.PP6138.2024
数据共享方式
开放下载
摘 要
Lithology identification of hydrocarbon reservoir based on well logging curve is significant for drilling planning and reservoir development. The conventional stratigraphic identification methods usually require large amount of labour work and thus time consuming. The machine learning is able to deal with the data-intensive tasks and enhancing the lithology/formation recognition, and make it easier to build a lithotype profile. Four typical algorithms, i.e. Adaboost, decision tree, random forest and linear support vector machine are introduced in present study. By comparing the prediction accuracy of each algorithm on TZ4 well logging data, an optimal algorithm is chosen for further study. Then, the sensitivity analysis of data dimensionality reduction and training set adjustment is carried out, so as to explore the main factors affecting the accuracy of formation identification. The results show that decision tree algorithm has higher accuracy over the selected four algorithms and when the training ratio reaches 50% and the number of logging parameters is at least four, the accuracy can reach 90%.
郭斯尧,叶智慧,王涵,陈冬,朱丹丹. Automatic Lithology Identification Based on Machine Learning: A Case Study of TZ4 Well（基于机器学习的岩性自动识别： TZ4 井案例研究）. 国家冰川冻土沙漠科学数据中心(http://www.ncdc.ac.cn), 2024. https://cstr.cn/CSTR:11738.14.NCDC.XDA14.PP6138.2024.
本作品采用
知识共享署名
4.0 国际许可协议进行许可。
CSTR:11738.14.NCDC.XDA14.PP6138.2024