钻井液与完井液2023，Vol.40Issue(5)：563-570,8.DOI:10.12358/j.issn.1001-5620.2023.05.003
基于可解释性机器学习的ECD敏感性分析与预测技术
ECD Sensitivity Analyses and Prediction Based on Interpretable Machine Learning
马磊 1周波 1张宁俊 2杨恒 3蔡新树 2刘征 2徐同台3
作者信息
- 1. 中国石油塔里木油田分公司,新疆库尔勒 841000
- 2. 昆仑数智科技有限责任公司,北京 102206
- 3. 北京石大胡杨石油科技发展有限公司,北京 102206
- 折叠
摘要
Abstract
The calculation of equivalent circulation density(ECD)of drilling fluids is very cumbersome and time-consuming,and the pattern of sensitivity of the ECD has not been quite understood yet.To solve this problem,1928 data from the Block Keshen were acquired and analyzed using the ProHydraulic software to calculate the theoretical ECD.Based on the calculation some characteristic parameters were determined.Key factors affecting the sensitivity of the ECDs of a drilling fluid,such as mud properties,drilling parameters and sizes of the annular spaces are analyzed using interpretable machine learning technology.Using linear regression,an empirical equation for calculating the ECDs of drilling fluids in the Block Keshen,which covers 12 main characteristic parameters,is constructed.ECD calculation with the model shows that the model gives results that fit the practical values with excellence.The coefficient of determination of the testing set is 0.963,and the average absolute error is only 0.04,indicating that this empirical equation is simple and efficient in ECD calculation in practical engineering application.关键词
ECD/敏感性分析/可解释性机器学习/线性回归模型Key words
ECD/Sensitivity analysis/Interpretable machine learning/Linear regression model分类
能源科技引用本文复制引用
马磊,周波,张宁俊,杨恒,蔡新树,刘征,徐同台..基于可解释性机器学习的ECD敏感性分析与预测技术[J].钻井液与完井液,2023,40(5):563-570,8.
