应用PCA和多元非线性回归快速预测储层敏感性

Application of PCA and multiple nonlinear regression to rapid prediction of reservoir sensitivity

王沛 欧阳传湘 陈宏生 陈向军

长江大学石油工程学院，湖北 武汉 430100 中国石化石油工程技术研究院，北京 100101

Petroleum Engineering College, Yangtze University, Wuhan 430100, China Research Institute of Petroleum Engineering, SINOPEC, Beijing 100101, China

通过对比分析近年来各种储层敏感性的预测方法，发现耦合PCA与多元非线性回归分析是一种快速、精确、泛化能力强的敏感性预测新方法。以盐敏为例，经过特征选择与提取，建立了预测储层盐敏损害指数的数学模型，并通过精度检验提高其泛化能力。借助塔里木油田的90组样本，检验了新模型在储层敏感性预测中的应用效果。结果表明，盐敏损害指数的平均准确率大于95%，从而证明耦合PCA与多元非线性回归的算法能达到快速、准确预测储层敏感性的目的。该方法操作简单，准确率高，泛化能力强，具有较好的应用前景。

Through comparison and analysis of the reservoir sensitivity prediction methods developed in recent years, it is found that the combination of PCA and nonlinear multiple regression is a new rapid and precise method in sensitivity prediction. In this paper the mathematics model is established by feature selection and extraction to predict the salty sensitive damage index, and the generalization ability is improved by precision test. The application of the method is verified by 90 groups of data in the Tarim Oilfield. The results show that the average precision of salt damage index is over 95%, which proves that the algorithm of combining PCA and nonlinear multiple regression is efficient in sensitivity prediction. This method is superior to others in veracity and generalization, and its application foreground is broad.

10.6056/dkyqt201802021