基于增强双向长短时记忆神经网络的测井数据重构  

Log data reconstruction method based on enhanced bidirectional long short-term memory neural network

作　　者：张海涛[1] 杨小明[1] 陈阵[1] 赵海航 周伟 黄兆辉[3] 赖富强[3] ZHANG HaiTao;YANG XiaoMing;CHEN Zhen;ZHAO HaiHang;ZHOU Wei;HUANG ZhaoHui;LAI FuQiang(Exploration and Development Research Institute of PetroChina Changqing Oilfield Company,Xi'an 710018,China;School of Intelligent Technology and Engineering,Chongqing University of Science and Technology,Chongqing 401331,China;School of Petroleum and Natural Gas Engineering,Chongqing University of Science and Technology,Chongqing 401331,China)

机构地区：[1]中国石油长庆油田分公司勘探开发研究院,西安710018 [2]重庆科技学院智能技术与工程学院,重庆401331 [3]重庆科技学院石油与天然气工程学院,重庆401331

出　　处：《地球物理学进展》2022年第3期1214-1222,共9页Progress in Geophysics

基　　金：国家科技重大专项“鄂尔多斯盆地大型岩性地层油气藏勘探开发示范工程”(2016ZX05050);重庆市教委科学技术研究项目“基于深度学习的碳酸盐岩储层缝洞自动识别及评价研究”(KJQN201801502)联合资助。

摘　　要：测井数据是油气田开发和评价的基础,对于地下储层地质建模和指导油气藏勘探具有重要意义.然而,由于实际开采过程中井径垮塌、仪器故障等原因往往导致整段甚至整条测井数据失真或缺失,而重新测井成本高昂,施工难度大.本文提出一种基于增强双向长短时记忆神经网络(简称:DBi-LSTM)的测井数据重构方法,在不增加额外测量成本的情况下,充分考虑缺失数据点的前趋和后继之间的双向关联性,通过增加双向长短时记忆神经网络深度来增强模型表达能力并利用迭代更新策略完成整段缺失数据块的重构.对比循环神经网络、长短时记忆神经网络及双向长短时记忆神经网络,实验表明提出的方法具有更好的数据重构精确度.Logging data is an important basis for the development and evaluation of oil and gas fields,which is of great significance for the geological modeling of low reservoir and guiding the exploration of oil and gas reservoirs.However,borehole collapse,tool failure and other reasons in the actual mining process often lead to the whole section or even the whole logging data distortion or missing.Unfortunately,the cost of re logging is high,and the construction is difficult.In this paper,a log data reconstruction method based on improved bi-directional long and short-term memory neural network is proposed.Without additional measurement cost,the correlation between missing data points and previous and subsequent data is fully considered,and the whole missing data block is reconstructed through iterative update strategy.Compared with Recurrent Neural Network(RNN),Long and Short-Term Memory neural network(LSTM)and Bidirectional Long and Short-Term memory neural network(Bi-LSTM),experiment results show that the proposed method has better reconstruction accuracy of logging data.

关 键 词：测井数据 重构 循环神经网络 双向长短时记忆神经网络 

分 类 号：P631[天文地球—地质矿产勘探]