基于叠前资料的新型属性优选算法

摘要

为了实现属性优选的定量化评价,提高属性优选的准确率,提出了一种新型属性优选算法,将非线性支持向量回归机(SVR)引入到遗传算法(GA)当中,在适应度评价时,使用SVR建立属性集与储层特征参数的定量计算关系,并且,首次将该新型属性优选算法应用到叠前叠后属性的优选.该方法在胜坨地区沙四纯上段进行应用,一方面避免了基于叠后地震属性的预测方法存在不确定性的问题,另一方面预测出了更加符合地质认识的储层展布结果.%In order to realize quantitative evaluation of attribute optimization and to improve the accuracy of attribute optimization,this paper puts forward a new algorithm of attribute optimization which introduces the nonlinear support vector regression(SVR) into genetic algorithm(GA) and uses SVR to build the quantitative calculation relationship between the attribute set and the reservoir characteristic parameters.In addition,the new algorithm of attribute optimization is the first time to be applied in attribute optimization of prestack and poststack.The new algorithm is used in the upper section of Sha4 in Shengtuo area,which not only avoids the uncertainty of prediction method based on poststack seismic attributes but also predicts reservoir distribution results which are more consistent with the geological laws.