基于自注意力机制生成对抗网络的三维储层建模方法

丁祖鹏; 张雨晴; 王俊杰; 方洪峰; 陈大颉; 陈麒玉

doi:10.19509/j.cnki.dzkq.tb20250109

您当前的位置：

首页 >

文章列表页 >

基于自注意力机制生成对抗网络的三维储层建模方法

地学信息技术与工程 | 更新时间：2026-11-05

- 基于自注意力机制生成对抗网络的三维储层建模方法
- 基于自注意力机制生成对抗网络的三维储层建模方法
- 地质科技通报 2025年44卷第4期页码：391-402
- 作者机构：
  
  1. 海洋油气高效开发国家重点实验室
  2. 中海油研究总院有限责任公司
  3. 中国地质大学（武汉）未来技术学院
  4. 中国地质大学（武汉）计算机学院
- 作者简介：
- 基金信息：
  
  国家自然科学基金项目（42172333）;中海石油（中国）有限公司北京研究中心横向协作项目（CCL2023RCPS0426RSN）
- DOI：10.19509/j.cnki.dzkq.tb20250109
  中图分类号： P618.13;TE34
- 网络首发：2025-06-26，
  
  纸质出版：2025-06-26
- 稿件说明：
移动端阅览
丁祖鹏, 张雨晴, 王俊杰, 等. 基于自注意力机制生成对抗网络的三维储层建模方法[J]. 地质科技通报, 2025,44(4):391-402.

丁祖鹏, 张雨晴, 王俊杰, et al. 基于自注意力机制生成对抗网络的三维储层建模方法[J]. Bulletin of Geological Science and Technology, 2025, 44(4): 391-402.
丁祖鹏, 张雨晴, 王俊杰, 等. 基于自注意力机制生成对抗网络的三维储层建模方法[J]. 地质科技通报, 2025,44(4):391-402. DOI： 10.19509/j.cnki.dzkq.tb20250109.

丁祖鹏, 张雨晴, 王俊杰, et al. 基于自注意力机制生成对抗网络的三维储层建模方法[J]. Bulletin of Geological Science and Technology, 2025, 44(4): 391-402. DOI： 10.19509/j.cnki.dzkq.tb20250109.

摘要

三维储层建模技术可实现储层空间异质结构的自动表征，然而部分油田开采难度大、开发成本高致使油田井距大、钻井资料少，如何根据稀疏有限的可用资料指导油气储层三维建模，一直是油气开发工作的难点。本研究提出了一种基于自注意力机制生成对抗网络的三维储层建模方法，引入具有丰富空间结构信息的地质剖面或钻孔作为条件约束数据，使用U-Net网络结构结合自注意力机制提取关键结构特征，设计空间上下文条件损失函数，以进一步约束重建过程，使得重构结果的条件分布更接近于真实数据的条件分布。多组三维地层结构及复杂砂岩孔隙建模实验结果表明，本研究提出的三维储层建模方法能够再现地质空间结构特征，且符合参考模型的条件数据分布，重构准确率达90%。本研究所提出的方法，成功捕捉了那些对于传统卷积层来说难以识别的远距离依赖特征，克服了由条件数据稀疏性引起的潜在问题，模拟结果可反映地质随机性，能够应用于多种储层地质结构的高效准确重建工作中。

Abstract

ObjectiveThree-dimensional reservoir modeling technology can automatically characterize the spatial heterogeneity structure of the reservoir. However

oilfield exploitation is difficult and the development cost is high

resulting in large well spacing and scarce drilling data. How to realize the three-dimensional modeling of oil and gas reservoirs based on the sparse and limited available data has always been a challenge in oil and gas development. MethodsTherefore

this paper proposes a three-dimensional reservoir modeling method based on the self-attention mechanism and generative adversarial network. Geological profiles or boreholes with rich spatial structure information are introduced as conditional constraint data. The U-Net network structure combined with the self-attention mechanism is used to extract key structural features. A spatial context conditional loss function is designed to further constrain the reconstruction process

so that the conditional distribution of the reconstructed result is closer to that of the real data. ResultsThe results of multiple sets of three-dimensional reservoir structure and complex sandstone pore modeling experiments show that the three-dimensional reservoir modeling method proposed in this study can reproduce the geological spatial structure features and is in line with the conditional data distribution of the reference model. The reconstruction accuracy rate is 90%. ConclusionsThe method proposed in this study successfully captures those long-distance dependent features that are difficult to identify for traditional convolutional layers. It overcomes the potential problems caused by the sparse condition data and the simulation results can reflect geological randomness. It can be applied to the efficient and accurate reconstruction of various reservoir geological structures.

关键词

Keywords

references

孙福街.中国海上油田高效开发与提高采收率技术现状及展望[J].中国海上油气,2023(05).

唐骥,蒋潇,姜雪莲,包金坤,姚丽香,席万鑫.矿体三维可视化建模技术在成矿模式分析中的应用[J].地质科技通报,2023(05).

曹砚锋,李汉兴,黄辉,程载斌,张磊,潘豪,邢希金,任革学,王利华,易会安,舒福昌,王超.海上油田高效开发钻完井关键技术研究新进展[J].中国海上油气,2022(06).

高小洋,何文祥,胡勇.基于深度生成模型的河道砂建模方法[J].地质科技通报,2023(02).

Tuan-Feng Zhang,Peter Tilke,Emilien Dupont,Ling-Chen Zhu,Lin Liang,William Bailey.Generating geologically realistic 3D reservoir facies models using deep learning of sedimentary architecture with generative adversarial networks[J].Petroleum Science,2019(03).

林良俊,李亚民,葛伟亚,胡秋韵,李晓昭,李云,孟晖,张礼中,杨建锋.中国城市地质调查总体构想与关键理论技术[J].中国地质,2017(06).

周飞燕,金林鹏,董军.卷积神经网络研究综述[J].计算机学报,2017(06).

吴冲龙,何珍文,翁正平,刘军旗.地质数据三维可视化的属性、分类和关键技术[J].地质通报,2011(05).

李德仁,龚健雅,邵振峰.从数字地球到智慧地球[J].武汉大学学报(信息科学版),2010(02).

Chen Dajie,Chen Qiyu,Cui Zhesi,Wang Ruyi,Liu Gang.SA-VAE: a novel approach for reservoir characterization based on variational auto-encoder and selective attention mechanism[J].Earth Science Informatics,2023.

Cui Zhesi,Chen Qiyu,Liu Gang.A two-stage downscaling hydrological modeling approach via convolutional conditional neural process and geostatistical bias correction[J].Journal of Hydrology,2023.

Cui Zhesi,Chen Qiyu,Liu Gang.Characterization of Subsurface Hydrogeological Structures With Convolutional Conditional Neural Processes on Limited Training Data[J].Water Resources Research,2022.

Chen Qiyu,Cui Zhesi,Liu Gang,Yang Zixiao,Ma Xiaogang.Deep convolutional generative adversarial networks for modeling complex hydrological structures in Monte-Carlo simulation[J].Journal of Hydrology,2022.

Yang Zixiao,Chen Qiyu,Cui Zhesi,Liu Gang,Dong Shaoqun,Tian Yiping.Automatic reconstruction method of 3D geological models based on deep convolutional generative adversarial networks[J].Computational Geosciences,2022.

Jessell Mark,Guo Jiateng,Li Yunqiang,Lindsay Mark,Scalzo Richard,Giraud Jérémie,Pirot Guillaume,Cripps Ed,Ogarko Vitaliy.Into the Noddyverse: a massive data store of 3D geological models for machine learning and inversion applications[J].Earth System Science Data,2022.

Cao Danping,Hou Zhiyu,Liu Qiang,Fu Feiqi.Reconstruction of three-dimension digital rock guided by prior information with a combination of InfoGAN and style-based GAN[J].Journal of Petroleum Science and Engineering,2022.

Cui Zhesi,Chen Qiyu,Liu Gang,Mariethoz Gregoire,Ma Xiaogang.Hybrid parallel framework for multiple-point geostatistics on Tianhe-2: A robust solution for large-scale simulation[J].Computers and Geosciences,2021.

Liu Qingbin,Liu Wenling,Yao Jianpeng,Liu Yuyang,Pan Mao.An Improved Method of Reservoir Facies Modeling Based on Generative Adversarial Networks[J].Energies,2021.

Erik Nesvold,Tapan Mukerji.Simulation of Fluvial Patterns With GANs Trained on a Data Set of Satellite Imagery[J].Water Resources Research,2021.

Lopez Alvis Jorge,Laloy Eric,Nguyen Frédéric,Hermans Thomas.Deep generative models in inversion: The impact of the generator's nonlinearity and development of a new approach based on a variational autoencoder[J].Computers & Geosciences,2021.

Goodfellow Ian,Pouget Abadie Jean,Mirza Mehdi,Xu Bing,Warde Farley David,Ozair Sherjil,Courville Aaron,Bengio Yoshua.Generative Adversarial Networks[J].COMMUNICATIONS OF THE ACM,2020.

Leonardo Azevedo,Gustavo Paneiro,Arthur Santos,Amilcar Soares.Generative adversarial network as a stochastic subsurface model reconstruction[J].Computational Geosciences,2020.

Qiyu Chen,Gang Liu,Xiaogang Ma,Xinchuan Li,Zhenwen He.3D stochastic modeling framework for Quaternary sediments using multiple-point statistics: A case study in Minjiang Estuary area, southeast China[J].Computers and Geosciences,2020.

Du Getao Cao Xu Liang Jimin Chen Xueli Zhan Yonghua.Medical Image Segmentation based on U-Net: A Review[J].Journal of Imaging Science and Technology,2020.

Department of Petroleum Engineering, University of Wyoming, Laramie, WY 82071, United States,Mork Family Department of Chemical Engineering and Materials Science, University of Southern California, Los Angeles, CA 90089-1211, United States,Department of Petroleum Engineering, University of Wyoming, Laramie, WY 82071, United States,Mork Family Department of Chemical Engineering and Materials Science, University of Southern California, Los Angeles, CA 90089-1211, United States.Machine learning in geo- and environmental sciences: From small to large scale[J].Advances in Water Resources,2020.

Department of Petroleum Engineering, University of Wyoming, Laramie, WY, 82071, USA,Department of Petroleum Engineering, University of Wyoming, Laramie, WY, 82071, USA.Hybrid geological modeling: Combining machine learning and multiple-point statistics[J].Computers and Geosciences,2020.

Renguang Zuo,Yihui Xiong,Jian Wang,Emmanuel John M. Carranza.Deep learning and its application in geochemical mapping[J].Earth-Science Reviews,2019.

Alessandro Comunian,Mauro Giudici,Luca Landoni,Sergio Pugnaghi.Improving Bowen-ratio estimates of evaporation using a rejection criterion and multiple-point statistics[J].Journal of Hydrology,2018.

Wenjie Feng,Shenghe Wu,Yanshu Yin,Jiajia Zhang,Ke Zhang.A training image evaluation and selection method based on minimum data event distance for multiple-point geostatistics[J].Computers and Geosciences,2017.

Mehrdad Soleimani,Behshad Jodeiri Shokri.3D static reservoir modeling by geostatistical techniques used for reservoir characterization and data integration[J].Environmental Earth Sciences,2015.

G. Burc Arpat,Jef Caers.Conditional Simulation with Patterns[J].Mathematical Geology,2007.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于不同条件模拟策略的生成对抗网络沉积相建模

基于深度学习的复杂页岩储层多参数协同高效预测——以松辽盆地青山口组为例

一种改进的LSTM横波预测方法：以塔河油田缝洞储层为例

基于深度学习和钻孔图像岩体RQD智能计算

基于微震事件频次的采空区沉陷变形智能预警方法