西安石油大学学报（自然科学版）2024，Vol.39Issue(1)：122-128,7.DOI:10.3969/j.issn.1673-064X.2024.01.015
基于LSTM神经网络深度序列机械钻速实时预测
Real-time Prediction of ROP Based on LSTM Neural Network Deep Sequence
摘要
Abstract
ROP is a key factor for optimizing drilling and shortening drilling cycles.Traditional ROP prediction mostly involves drilling analysis after drilling,with low prediction efficiency and accuracy,and limited applicability to geological formations.In order to efficient-ly predict high-precision ROP,a deep sequence ROP prediction method based on Long Short Term Memory(LSTM)neural network is proposed.Collect real-time drilling datasets,select 8 parameters including well depth,gamma rays,formation density,pore pressure,wellbore diameter,drilling time,drilling displacement,and drilling fluid density as characteristic parameters,and use Pearson correlation coefficient to measure the correlation between characteristic parameters.Build an LSTM neural network model,train the LSTM model and predict ROP,and analyze the prediction results.The performance of LSTM model,BP model,and SVM model is comparatively ana-lyzed by means of determination coefficient(R2),root mean square error(RMSE)and mean absolute percentage error(MAPE).The results show that the R2,RMSE and MAPE of LSTM model are 0.948、1.151 and 17.075.Compared to BP model and SVM model,the R2 of LSTM model is smaller and the RMSE and MAPE are larger,which shows better predictive performance of the LSTM model.This method helps drilling engineers and decision-makers to obtain drilling information in advance,thereby better planning drilling opera-tions,shortening drilling cycles,and providing a new way to predict drilling parameters.It can improve the efficiency and accuracy of previous prediction methods in dealing with complex geological problems.关键词
机械钻速/LSTM神经网络/深度序列/实时预测/人工智能/深度学习Key words
ROP/LSTM neural network/deep sequence/real-time prediction/artificial intelligence/deep learning分类
能源科技引用本文复制引用
冯义,朱亮,杨立军,李慎越,席俊卿,陈芳,纪慧..基于LSTM神经网络深度序列机械钻速实时预测[J].西安石油大学学报（自然科学版）,2024,39(1):122-128,7.基金项目
中国石油天然气股份有限公司前瞻性基础性技术攻关项目"致密气勘探开发技术研究"(2022DJ2107) （2022DJ2107）
