Jake south油田AG组油藏地质条件极其复杂,内部存在大量的低电阻率油层,识别难度大,准确率不高.现有模式识别算法常采用的部分模型变量不能反映储层流体性质,在地质条件复杂的油藏中应用效果往往不好.综合岩心、薄片、扫描电镜、测井曲线等资料,对模式识别中通常所选变量进行优化,并用于低电阻率油层识别中.结果表明,储层类型、自然伽马、自然电位幅度比、中子密度指数、原状地层电阻率、声波时差、补偿中子、地层密度等变量可反映储层流体性质,建议采用.采用支持向量机方法,选用建议的模型变量对已证实油水层进行识别,取得良好效果,油水层识别准确率达94.4%,远高于采用以往常用变量的准确率.%The geological conditions of the reservoirs in AG Formation are very complicated.Many low resistivity pays are developed,the identification process is difficult and the accuracy is low.The model variables are always used by the pattern recognition methods,such as support vector machine and neural networks that are incapable to reflect fluid types and the model accuracy is low.Cores,thin sections,SEM and well logs are comprehensively utilized to choose better model variables for the pattern recognition methods in low resistivity pay identification.The results show that reservoir types,natural gamma,the ratio of spontaneous potential amplitude RSP,neutron-density index ICD,deep resistivity,acoustic time,compensate neutron and balk density are recommended as the model variables.The suggested variables and the variables are often used in support vector machine for fluid identification of the proved reservoirs in the studied area.The results show the accuracy with suggested variables reaches 94.4%,which is significantly higher than that with the variables often used.
展开▼
