Interpretation and characterization of rate of penetration intelligent prediction model

作     者：Zhi-Jun Pei Xian-Zhi Song Hai-Tao Wang Yi-Qi Shi Shou-Ceng Tian Gen-Sheng Li Zhi-Jun Pei;Xian-Zhi Song;Hai-Tao Wang;Yi-Qi Shi;Shou-Ceng Tian;Gen-Sheng Li

作者机构：College of Petroleum EngineeringChina University of Petroleum(Beijing)Beijing102249China State Key Laboratory of Petroleum Resources and ProspectingBeijing102249China Kunlun Digital Technology Co.Ltd.Beijing102206China 

出 版 物：《Petroleum Science》 (石油科学（英文版）)

年 卷 期：2024年第21卷第1期

页      面：582-596页

核心收录：

学科分类：0820[工学-石油与天然气工程] 08[工学] 082002[工学-油气田开发工程] 

基　　金：The authors greatly thanked the financial support from the National Key Research and Development Program of China(funded by National Natural Science Foundation of China,No.2019YFA0708300) the Strategic Cooperation Technology Projects of CNPC and CUPB(funded by China National Petroleum Corporation,No.ZLZX2020-03) the National Science Fund for Distinguished Young Scholars(funded by National Natural Science Foundation of China,No.52125401) Science Foundation of China University of Petroleum,Beijing(funded by China University of petroleum,Beijing,No.2462022SZBH002) 

主　　题：Fully connected neural network Explainable artificial intelligence Rate of penetration ReLU active function Deep learning Machine learning 

摘      要：Accurate prediction of the rate of penetration(ROP)is significant for drilling *** the intelligent ROP prediction model based on fully connected neural networks(FNN)outperforms traditional ROP equations and machine learning algorithms,its lack of interpretability undermines its *** study proposes a novel interpretation and characterization method for the FNN ROP prediction model using the Rectified Linear Unit(ReLU)activation *** leveraging the derivative of the ReLU function,the FNN function calculation process is transformed into vector *** FNN model is linearly characterized through further simplification,enabling its interpretation and *** proposed method is applied in ROP prediction scenarios using drilling data from three vertical wells in the Tarim *** results demonstrate that the FNN ROP prediction model with ReLU as the activation function performs exceptionally *** relative activation frequency curve of hidden layer neurons aids in analyzing the overfitting of the FNN ROP model and determining drilling data *** the well sections with similar drilling data,averaging the weight parameters enables linear characterization of the FNN ROP prediction model,leading to the establishment of a corresponding linear representation ***,the quantitative analysis of each feature s influence on ROP facilitates the proposal of drilling parameter optimization schemes for the current well *** established linear characterization equation exhibits high precision,strong stability,and adaptability through the application and validation across multiple well sections.