最小二乘支持向量机在储层流体识别中的应用

Application of Least-square Support Vector Machine Method in Identifying Reservoir Fluids

在测井储层流体识别中引入基于统计学习理论的最小二乘支持向量机(LS-SVM)算法,它是在传统的支持向量机(SVM)基础之上加以改进的一种新算法。LS-SVM采用结构风险最小化原则代替了传统的经验风险最小化原则,保证了其具有全局最优性和较好的泛化能力,并且它将凸二次规划问题转变成了线性方程组的求解问题,使计算效率大大提高。介绍了LS-SVM方法的基本原理和多分类方法,通过该法利用少量的测井资料作为学习样本,准确地对油气水层进行了识别。将它与交会图判别法和BP神经网络方法的预测结果进行比较,表明用LS-SVM方法来进行储层流体识别是可行的,且具有一定的优越性。

The method of least -square support vector machine (LS-SVM),which was an improved algorithm for traditional support vector machine(SVM) based on statistical learning theory,was presented to study the recognition of reservoir fluids from geologic logging. By using the principle of structure risk minimization as opposed to empirical risk supported by conventional technique,the LS-SVM method was very good in global optimization and generalization. Furthermore,it acquires a high computing efficiency by solving a set of linear equations instead of solving a convex quadratic programming problem. It can be used to introduce the basic principle and multi-classified method of LS-SVM algorithm,then make use of a small quantity of well logging information as training sample to recognize the oil and water zones accurately based on this method. The comparison of predicted classification with the result by employing cross plot and BP neural network methods shows that the proposed method of LS-SVM is not only feasible in correctly recognizing the fluids of reservoir,but also gains much superiority.