线性回归理论在多属性预测储层参数中的应用

Application of Linear Regression Theory In Multi-attributes Reservior Parameter Prediction

讨论了一种非常规的反演方法,通过建立地震属性与测井曲线之间的统计模型——多元线性回归模型,预测各种储层参数。在模型建立过程中,用交互有效性来检验模型的可信度,并且采用逐步回归的方法来确定属性选取的优先权。应用该方法,理论上可以预测储层物性参数的分布。

Instead of assuming a conventional seismic acoustic impedance inversion model relating the logs and the seismic data,a statistical relationship was derived by analyzing a set of seismic attributes data and well logs at well locations.This relationship is linear-multivariate regression.The multivariate regression is used to predict log properties,the concept of crossvalidation is used to estimate the reliability of the derived relationship and use step-wise regression to determine the priority of seismic attributes.By using this method,the distribution of reservoir physical property can be predicted.