基于统计学习理论的高含盐油藏储层渗透率变化预测  

Prediction of permeability in highly saliferous oil reservoir based on statistical learning theory

作　　者：尤启东[1] 陈月明[1] 

机构地区：[1]中国石油大学(华东)石油工程学院,山东东营257061

出　　处：《油气地质与采收率》2006年第2期75-77,共3页Petroleum Geology and Recovery Efficiency

基　　金：中国石化基金项目"王场油田水驱储层参数变化机理与规律研究"(P03056)部分成果

摘　　要：为了在较少的实验数据条件下,实现对高含盐油藏储层渗透率变化规律的有效预测,对自组织、改进型BP神经网络和支持向量机3种方法在水驱储层渗透率变化预测中的应用进行了探讨。3种方法的对比研究表明,在小样本条件下,支持向量机方法能够兼顾模型的通用性和推广性。在王场油田潜三段北断块油藏储层渗透率变化的敏感性分析应用结果表明,该方法可准确地预测储层渗透率的变化规律;编制的动态油藏数值模拟软件应用结果显示,考虑储层渗透率变化的剩余油数值模拟结果符合率达75%,而不考虑储层渗透率变化的结果符合率仅为45%,充分说明了动态模拟的优越性。In order to effectively predict the changing rule of permeability in the highly saliferous oil reservoir with fewer experimental data, the applications of group method of data handling (GM-DH) ,improved error back propagation (BP) artificial neutral network and support vector machine (SVM) to the prediction of permeability change in water drive reservoir were discussed. The comparison among the three methods indicates that SVM has both the universality and the popularity when the samples are very limited for a model. The application of this method to the sensibility analysis of the permeability change in north block reservoir in Qian3 member in Wangchang Oilfield indicated that this method can predict the changing rule of the permeability in the saliferous reservoir accurately. The results of the application of the programmed dynamic reservoir numerical simulation software show that the 75% of the result considering the change of permeability was correct, while only 45% of that without considering the change. Dynamic numerical simulation shows a good prospect.

关 键 词：统计学习理论 BP神经网络 支持向量机 渗透率 预测 高含盐油藏 

分 类 号：TE319[石油与天然气工程—油气田开发工程]