支持向量机方法在储层预测中的应用

传统储层预测学习方法大都基于经验风险最小化准则，预测效果不理想。而基于结构化风险最小化准则的支持向量机方法，通过对推广误差（风险）上界的最小化达到最大的泛化能力和全局最优，具有可靠的预测能力。对支持向量机法的方法原理，即非线性模式识别法和非线性函数估计法进行了讨论，并采用不同的样本数，将其与神经网络法作对比，结果表明，2种方法的训练结果精度都较高，但对sinc函数的估计结果，支持向量机法更可靠。在胜利油田某区块应用了向量机法，以地震波波形作为输入向量进行了砂体孔隙度和含油性预测，预测结果与已知结果吻合较好。

Application of SVM method in reservoir prediction

Most classical learning methods are based on the empirical risk minimization (ERM) rule. Usually, these methods exist an over fitting problem when being used to resolve actual problem. By generalizes the error topside's minimization, the Support Vector Machine (SVM) method namely are nonlinear pattern recognition method and nonlinear function estimation method based on the structure risk minimization can get maximum universality and global optimizatioa Using a sandy body in Shengli Oil field as a research target, waveform datum is taken as input vectors. This method makes use of seismic waveform's characters completely. At the same time, it avoids plenty of works during attributes optimization and abstracting parameters partly. This way can be carried out more conveniently and shows a good result.