能够同时对多种属性进行训练，具有优秀推广能力的支持向量机（Support Vector Machine，简称SVM）方法是进行高精度地震参数预测的有力保障。然而，支持向量机中用于构建回归估计函数的参数最优解很难确定。针对该问题，通过建立数学模型进行参数选择研究，总结出了参数ε、C、σ2对样本预测的影响规律。在此基础上提出了求取惩罚因子C和核参数σ2的权系数公式。结合提出的参数求取公式，利用支持向量机方法，以地震属性为输入向量对渤海SZ36-1油田的砂泥岩百分比和孔隙度进行了预测。结果表明，利用该方法对储层参数进行预测具有较高的预测精度；权系数公式的提出极大地缩短了构建回归估计函数所耗用的时间，简化了参数选取的难度。%Having the ability to train on multi-attribute and provided with excellent generalization performance simultaneously, the sup-port vector machine (SVM) method is a powerful guarantee in high precision seismic parameters prediction. However, it is difficult to determine optimal solution of parameters for constructing regression estimation function in support vector machine. In allusion to the is-sue, through mathematical modeling studied parameter selection, summed up impact pattern from parameters ε, C andσ2 on sample prediction. On this basis has put forward weighting coefficient formula to get penalty factor C and kernel parameterσ2. Combined with parameter getting formula, using support vector machine method has carried out prediction of sandstone-mudstone percentage and po-rosity in SZ36-1 oilfield. The result has shown:using the method carry out reservoir parameter prediction has rather high prediction ac-curacy. The proposed weighting coefficient formula has greatly reduced regression estimation function constructing time, and simplified difficulty in parameter selection.
展开▼
