[1]韩 建,陈 着,王业统,等.基于RGCVAE的测井曲线重构方法[J].工业仪表与自动化装置,2025,(05):87-91.[doi:10.19950/j.cnki.CN61-1121/TH.2025.05.016]
　HAN Jian,CHEN Zhuo,WANG Yetong,et al.Reconstruction method of logging curves based on RGCVAE[J].Industrial Instrumentation & Automation,2025,(05):87-91.[doi:10.19950/j.cnki.CN61-1121/TH.2025.05.016]

点击复制

基于RGCVAE的测井曲线重构方法()

《工业仪表与自动化装置》[ISSN:1000-0682/CN:61-1121/TH]

卷:

期数:

2025年05期

页码:

87-91

栏目:

出版日期:

2025-10-15

文章信息/Info

Title:

Reconstruction method of logging curves based on RGCVAE

文章编号:

1000-0682（2025）05-0087-05

作者:

韩 建1; 3; 陈 着1; 3; 王业统2; 曹志民1; 3; 邓 宇1; 3

1.东北石油大学 三亚海洋油气研究院, 海南 三亚 572000;2.海南科技职业大学 虚拟现实技术与系统海南省工程研究中心, 海南 海口 571126;3.东北石油大学 物理与电子工程学院, 黑龙江 大庆 163318

Author(s):

HAN Jian1; 3; CHEN Zhuo1; 3; WANG Yetong2; CAO Zhimin1; 3;  DENG Yu1

1. Northeast Petroleum University Sanya Institute of Marine Oil and Gas Research, Hainan Sanya 572000, China; 2. Hainan Vocational University of Science and Technology Virtual Reality Technology and System Hainan Provincial Engineering Research Center, Hainan Haikou 571126, China;3. School of Physics and Electronic Engineering, Northeast Petroleum University, Heilongjiang, Daqing 163318, China

关键词:

循环格兰杰变分编码器; 测井曲线; 重构方法; 声波时差曲线

Keywords:

recurrent granger causality variational autoencoder;  logging curve;  reconstruction method;  acoustic time difference curve

分类号:

TP391

DOI:

10.19950/j.cnki.CN61-1121/TH.2025.05.016

文献标志码:

A

摘要:

在实际测井过程中，测井曲线的质量常常受到仪器故障和环境因素的影响，导致测井数据出现缺失。该文提出了一种基于RGCVAE的测井曲线重构方法，并结合大庆油田古工业区和金工业区的实际测井数据，分别进行了同井间和异井间的缺失数据重构实验。通过与随机森林、RNN和LSTM网络的实验结果进行对比分析，结果表明，RGCVAE模型在预测精度方面表现较好。在同井实验中，两口井重构后的声波时差曲线原始曲线的相关性分别达到了90.94%和88.60%；在异井实验中，两口井重构后的声波时差曲线与原始曲线的相关性分别为87.85%和85.71%。

Abstract:

In the actual logging process, the quality of logging curves is often compromised by instrument malfunctions and environmental factors, leading to missing logging data. This paper introduces a logging curves reconstruction method based on recurrent granger causality variational autoencoder. By utilizing actual logging data from the ancient and gold industrial areas of Daqing Oilfield, we conducted missing data reconstruction experiments for both intra-well and inter-well scenarios. Through comparative analysis with experimental results obtained from random forest, recurrent neural network, and long short-term memory networks, it was found that the recurrent granger causality variational autoencoder model outperforms other methods in terms of prediction accuracy. In the intra-well experiment, the correlation of the reconstructed acoustic moveout curve and the original curve of the two wells reached 90.94% and 88.60%, respectively. In the inter-well experiment, the correlation between the reconstructed acoustic time difference curve and the original curve of the two wells is 87.85% and 85.71%.

备注/Memo

备注/Memo:

收稿日期：2025-03-03第一作者：韩建（1976—），男，汉族，黑龙江大庆人，博士，教授，研究方向油井信号检测，机器学习，模式识别。

常用功能