【摘要】 提出了基于SOM(自组织神经网络)的无井约束剩余油分布预测方法:利用RBF神经网络预测时移前的原始含油饱和度;利用SOM对含油饱和度敏感的多属性差异进行聚类;利用动态资料和岩石物理测试数据圈定各种变化区域,建立各个区域含油饱和度变化与多属性差异之间的关系,依据所建立的关系获得储层含油饱和度的变化;将时移前的原始含油饱和度与含油饱和度变化相减得到储层的剩余油分布。该方法在渤海S油田取得了比较好的应用效果。更多还原

【Abstract】 A method to predict remaining oil distribution without drilled-well constraints has been developped on a basis of SOM(self-organization neural network).In this prediction,the original oil saturation before the time lapse is first predicted by using RBF neural networks,and then a cluster analysis is conducted for the multi-attribute differences sensitive to oil saturation by SOM.Thirdly,a relationship between oil saturation and multi-attribute difference can be established for various areas delineated by reservoir performance data and petrophysical data,and according to this relationship,the oil saturation change in reservoirs can be determined.Finally,the original oil saturation before the time lapse minus the oil saturation change would result in the remaining oil distribution in reservoirs.This method has led to better application results in Bohai S oilfield.更多还原