非线性支持向量机在油气预测中的应用

NONLINEAR SUPPORT VECTOR MACHINE AND ITS APPLICATION TO HYDROCARBON PREDICTION IN OIL AND GAS EXPLORATION

油气预测的传统方法通常是基于经验风险最小化准则,但在有限样本情况下,预测效果并不理想。研究引入基于结构化风险最小化准则的非线性支持向量机方法,通过对推广误差界的最小化达到最大的泛化能力和全局最优,对于小样本数据,该方法具有可靠的预测能力。在对四川观音场构造面新统上部碳酸盐岩储层数据处理中,通过实例试算,结果表明该方法有效可靠,预测精度高,与已知结果吻合较好。

The methods based on empirical risk minimization are often applied to hydrocarbon discrimination in oil and gas exploration. But the predictive validities of these methods are not perfect with small sample data. This paper introduces a nonlinear support vector machine （SVM） based on structural risk minimization. It obtains global optimization other than local one and with a better generalization. The nonlinear SVM is with robust predictive performance, especially in small samples. Experimental results in small data show that the nonlinear SVM is robust and may obtain higher recognition rates. This method is effective in hydrocarbon detection or discrimination in reservoir prediction of carbonate rocks.