中国石油大学学报（自然科学版）2023，Vol.47Issue(6)：50-59,10.DOI:10.3969/j.issn.1673-5005.2023.06.006
深度学习与Eaton法联合驱动的地层孔隙压力预测方法
A novel prediction method of formation pore pressure driven by deep learning and Eaton method
摘要
Abstract
The accurate prediction of formation pore pressure in deep complex formations has been one of the challenges in drilling engineering.In this paper,the limitations of the traditional Eaton methods and the shortcomings of the existing data-driven methods were discussed,and a convolutional neural network(CNN)and short-term memory network(LSTM)combi-nation model was constructed to fully explore the complex nonlinear relationship between drilling and recorded multi-source data and Eaton index by constructing an extension method for measuring formation pressure points.Based on the limited for-mation pressure data measured in drilled formations,a precise prediction of the Eaton index of the entire well can be a-chieved,which can provide an effective means for accurate prediction of the formation pore pressure in new wells with few measured points and uneven distribution of formation pressure.Field case studies show that the average relative error of the method established for predicting pore pressure in deep complex formations is 2.70%,while the average relative error of the traditional Eaton and LSTM methods is 7.60%and 5.12%,respectively.The combination model of deep learning with Eaton method,not only can improve the prediction accuracy of deep complex formation pore pressure,but it can also integrate multi-source data response features into the traditional methods,providing a theoretical support for the data-driven methods.关键词
Eaton法/数据驱动/深度学习/地层孔隙压力Key words
Eaton method/data-driven/deep learning/formation pore pressure分类
能源科技引用本文复制引用
许玉强,何保伦,王䶮舒,韩超,肖凡,管志川,刘宽..深度学习与Eaton法联合驱动的地层孔隙压力预测方法[J].中国石油大学学报（自然科学版）,2023,47(6):50-59,10.基金项目
国家自然科学基金面上项目(52074326) （52074326）
山东省优秀青年科学基金项目(ZR2023YQ045) （ZR2023YQ045）
