论文部分内容阅读
利用测井资料预测渗透率,传统的多元回归法需要先假定一个函数关系。由于物性变量之间的关系很难确切把握,所以一直不可能预先确定所采用的自变量和因变量之间潜在的函数形式。在石油地质特征变化较大将别是变量较多的情形下,参数回归常常失败,从而导致结论的不稳定甚至是错误的。 本文介绍了一种非参数方法,该方法通过对物性数据作最佳变换估算从而使观查变量间达到最大相关。这种方法不需要事先假定一个函数形式,其最佳变换仅依赖于数组。不象神经网络,其变换方便机械地依赖于函数识别。迭代过程涉及的交替条件期望(ACE)是该方法的基础。ACE的优势将用假设例子和油田实例加以说明。例子的结果清楚地显示了用ACE法提高渗透率预测与传统的参数回归法之间的比较。
Using log data to predict permeability, the traditional multiple regression method needs to assume a functional relationship. Since it is hard to know exactly the relationship between the physical variables, it has not been possible to determine in advance the potential functional forms of the dependent and dependent variables used. In the case of large changes in the characteristics of petroleum geology, the variable regression will often fail, leading to instability or even wrong conclusions. This paper presents a nonparametric method that maximizes the correlation between observed variables by making the best possible estimate of the physical data. This method does not need to assume a form of function, the best transformation depends only on the array. Unlike neural networks, its transformation is mechanically dependent on function recognition. The iterative process involves alternating conditions of expectations (ACE) is the basis of the method. The advantages of ACE will be demonstrated using hypothetical examples and field examples. The results of the example clearly show a comparison between the predictions of permeability enhancement by the ACE method and the traditional regression of parameters.