西安石油大学学报（自然科学版）2025，Vol.40Issue(3)：48-56,9.DOI:10.3969/j.issn.1673-064X.2025.03.005
基于深度学习的起下钻波动压力预测方法
Prediction Method of Pulling-out and Going-down Fluctuating Pressure during Drilling Based on Deep Learning
摘要
Abstract
In pulling-out and going-down operations,controlling the fluctuation pressure inside the well is crucial for maintaining the balance of wellbore pressure.However,the measurement of bottomhole pressure faces problems such as data transmission limitations,in-sufficient accuracy and poor timeliness of mechanism model calculations.Based on a large amount of historical data on pulling-out and going-down operations,the coupling relationship between fluctuating pressure and construction parameters was discussed,and a predic-tion method of pulling-out and going-down fluctuating pressure during drilling based on deep-learning was proposed.Through the exam-ple of offshore oil wells,the effectiveness of different fluctuation pressure prediction models was compared and analyzed.It was shown that the average absolute error of the DNN model in predicting the fluctuation pressure was 0.108 7 MPa,and the correlation coefficient R2 was 0.997 4,which verified the feasibility of this method and provided theoretical reference for safe and rapid pulling-out and going-down operations during drilling.关键词
波动压力/钻井液流变性/XGBoost/深度学习网络/起下钻Key words
fluctuation pressure/rheological properties of drilling fluid/XGBoost/deep learning network/drilling pulling-out and going-down operation分类
石油、天然气工程引用本文复制引用
魏刚,张晓广,李文良,夏环宇,高耸..基于深度学习的起下钻波动压力预测方法[J].西安石油大学学报（自然科学版）,2025,40(3):48-56,9.基金项目
中海油能源发展股份有限公司科技重大专项"监督业务数智化关键技术研究"(KFKJ-ZX-GJ-2023-01) （KFKJ-ZX-GJ-2023-01）
