摘要
由于地层结构的复杂性和强非均质性,同时受到测井、岩心、试油等数据不足的影响,现有沉积微相建模方法难以实现精确建模。提出一种基于条件生成对抗网络的沉积微相建模方法,采用灰色关联分析算法,计算各地震属性与砂地比的灰色关联度,挖掘对砂地比参数关联性较强的参数;将优选地震属性图像作为卷积神经网络模型的输入,构建砂地比预测模型,可视化砂地比预测结果,与井相图作为联合约束条件,训练条件生成对抗网络,构建沉积微相生成模型,实现沉积微相的精确建模。应用本方法对东部某油田进行沉积微相建模研究。结果表明,条件生成对抗网络沉积微相模型能精确刻画复杂地质模式,井点吻合率达到94.1%。
Due to the complexity and strong heterogeneity of stratigraphic structure,as well as the limited availability of logging,core,and oil testing data,existing sedimentary microfacies modeling methods struggle to achieve accurate results.To address this challenge,a new modeling approach based on conditional generative adversarial networks(cGANs)was proposed.This method utilizes grey correlation analysis to calculate the degree of correlation between various seismic attributes and the sand-to-ground ratio,thereby identifying attributes with strong predictive relevance.These selected seismic attribute images are then used as inputs to a convolutional neural network,which is employed to construct a prediction model for the sand-to-ground ratio.The resulting predictions are visualized as a thermal map,which,combined with well log phase diagrams,serves as a joint constraint for training the generative adversarial network.Based on this,a sedimentary microfacies generation model is developed to enable accurate modeling of sedimentary microfacies.This method was applied to a case study of an oilfield in eastern China.The results demonstrate that the cGAN-based model can effectively capture complex geological patterns,achieving a well-point coincidence rate of 94.1%.
作者
刘昕
孙胜
张立强
蔡明俊
鲁玉
卢文娟
LIU Xin;SUN Sheng;ZHANG Liqiang;CAI Mingjun;LU Yu;LU Wenjuan(Qingdao School of Software and School of Computer Science and Technology,China University of Petroleum(East China),Qingdao 266580,China;PetroChina Dagang Oilfield Company,Tianjin 300380,China)
出处
《中国石油大学学报(自然科学版)》
北大核心
2025年第4期1-10,共10页
Journal of China University of Petroleum(Edition of Natural Science)
基金
山东省自然科学基金项目(ZR2024MF037)。
关键词
条件生成对抗网络
深度学习
沉积微相
砂地比
灰色关联
卷积神经网络
conditional generative adversarial network
deep learning
sedimentary microfacies
sand-to-ground ratio
grey correlation
convolutional neural network
