基于主动学习代理集合辅助的油藏生产优化新方法  

A novel surrogate ensemble assisted reservoir production optimization method based on active learning

作　　者：沈子翔 王链 徐振平[3] SHEN Zixiang;WANG Lian;XU Zhenping(School of Automation,Central South University,Changsha 410083,China;ZhenHua Oil Co.,Ltd.,Beijing 100031,China;School of Computer Science,Yangtze University,Jingzhou 434000,China)

机构地区：[1]中南大学自动化学院,湖南长沙410083 [2]振华石油控股有限公司,北京100031 [3]长江大学计算机科学学院,湖北荆州434000

出　　处：《陕西理工大学学报(自然科学版)》2025年第2期22-29,共8页Journal of Shaanxi University of Technology:Natural Science Edition

基　　金：长江大学研究生教材建设立项项目(YJC202102)。

摘　　要：基于代理模型的优化方法能够实现每个生产阶段对注采方案进行高效动态调整,获得更高的油藏开发采收率和经济效益。现有代理优化方法需要大量计算昂贵的样本来构建代理模型。为此,提出了一种基于主动学习策略和代理集合的代理辅助多目标油藏生产优化方法(ALSA-MOPO)。该方法采用3种常用的代理模型径向基函数网络、高斯过程回归和支持向量回归来构建代理集合,并利用主动学习策略减少建立代理模型所需的样本数量,采用粒子群优化算法来搜索代理集合中最具不确定性和最优的样本,提升代理集合的质量和准确性。在Egg模型上的验证结果表明,在相同迭代次数下,ALSA-MOPO方法在收敛性和多样性方面均优于传统方法,将优化效率提高了50倍,实现了油藏生产的快速决策。The optimization method based on the proxy model can efficiently and dynamically adjust the injection and production plan at each production stage,achieving higher oil recovery rates and economic benefits in reservoir development.The existing proxy optimization methods require a large number of computationally expensive samples to construct proxy models.To address this issue,this paper proposes an agent assisted multi-objective reservoir production optimization method based on active learning strategy and agent set(ALSAMOPO).The ALSA-MOPO method uses three commonly used surrogate models:radial basis function network,Gaussian process regression,and support vector regression to construct surrogate sets,and utilizes active learning strategies to reduce the number of samples required to establish surrogate models.Particle swarm optimization algorithm is used to search for the most uncertain and optimal samples in the surrogate set,improving the quality and accuracy of the surrogate set.The validation results on the Egg model show that the ALSAMOPO method outperforms traditional methods in terms of convergence and diversity with the same number of iterations,increasing optimization efficiency by 50 times and achieving rapid decision-making for reservoir production.

关 键 词：主动学习 集合代理 QBC更新策略 多目标优化 生产优化 

分 类 号：TE341[石油与天然气工程—油气田开发工程]