基于深度生成网络的时变井控下油藏动态预测代理模型  

Surrogate model for reservoir performance prediction with time-varying well control based on depth generative network

作　　者：李艳春 贾德利[2] 王素玲[1] 屈如意 乔美霞 刘合[2,3] LI Yanchun;JIA Deli;WANG Suling;QU Ruyi;QIAO Meixia;LIU He(College of Mechanical Science and Engineering,Northeast Petroleum University,Daqing 163318,China;PetroChina Research Institute of Petroleum Exploration&Development,Beijing 100083,China;National Key Laboratory of Green Exploitation of Continental Shale Oil with Multi-Resource Collaboration,Daqing 163712,China)

机构地区：[1]东北石油大学机械科学与工程学院,黑龙江大庆163318 [2]中国石油勘探开发研究院,北京100083 [3]多资源协同陆相页岩油绿色开采全国重点实验室,黑龙江大庆163712

出　　处：《石油勘探与开发》2024年第5期1114-1125,共12页Petroleum Exploration and Development

基　　金：国家自然科学基金基础科学中心项目“数字经济时代的资源环境管理理论与应用”(72088101);国家自然科学基金面上项目“大数据驱动下的老油田水驱精细分析方法研究”(52074345);国家自然科学基金面上项目“非均质强塑性陆相页岩油储层水力复杂裂缝扩展机理研究”(52274036)。

摘　　要：在流形空间内定义并求解油藏动态预测问题,充分考虑地质不确定性和随时间变化的井控条件(简称时变井控)下油藏动态的变化特性,构建基于条件演化生成对抗网络(CE-GAN)的油藏动态预测代理模型。CE-GAN通过特征空间的条件演化使原来无法控制方向的生成网络实现定向演化,将油藏动态预测问题转化为基于渗透率分布、初始油藏动态和时变井控的图像演化问题,实现时变井控条件下油藏动态的快速准确预测。基础油藏模型(Egg模型)与实际油藏模型的验证结果表明,CE-GAN预测与数值模拟结果的一致性较好,基础油藏模型验证中压力和含油饱和度的相对残差中位数分别为0.5%和9.0%,实际油藏模型验证中压力和含油饱和度相对残差中位数均为4.0%;CE-GAN代理模型训练完成后,相较于传统数值模拟,分别将基础油藏模型和实际油藏模型的计算速度提升约160倍和280倍,可以有效提高生产优化的效率。This paper proposes a novel intelligent method for defining and solving the reservoir performance prediction problem within a manifold space,fully considering geological uncertainty and the dynamic characteristics of reservoirs under time-varying well control conditions,creating a surrogate model for reservoir performance prediction based on Conditional Evolutionary Generative Adversarial Networks(CE-GAN).The CE-GAN leverages conditional evolution in the feature space to direct the evolution of the generative network in previously uncontrollable directions,and transforms the problem of reservoir performance prediction into an image evolution problem based on permeability distribution,initial reservoir performance and time-varying well control,thereby enabling fast and accurate reservoir performance prediction under time-varying well control conditions.The experimental results in basic(egg model)and actual water-flooding reservoirs show that the model predictions align well with numerical simulations.The median of relative residuals of pressure and oil saturation for the basic reservoir model are 0.5%and 9.0%,respectively,while those for the actual reservoir model are both 4.0%.Regarding time efficiency,the surrogate model after training achieves approximately 160-fold and 280-fold increases in computational speed for the basic and actual reservoir models,respectively,compared with traditional numerical simulations.The reservoir performance prediction surrogate model based on the CE-GAN can effectively enhance the efficiency of production optimization.

关 键 词：深度生成网络 代理模型 时变井控 水驱开发 油藏动态 

分 类 号：TE341[石油与天然气工程—油气田开发工程]