测井序列变点分析自动识别岩性界面

A NEW APPROACH TO AUTOMATED IDENTIFICATION OF LITHOLOGIC BOUNDARIES: APPLICATION OF CHANGE-POINT ANALYSIS TO GEOPHYSICAL WELL LOG DATA

将非线性统计学中的变点分析方法引入到地球物理测井序列分析中,实现了自动岩性界面识别和自动分层。以自然伽马测井曲线为例,采用均值变点模型,用最小二乘法确定测井序列的变点,将其解释为岩性变化界面。根据测井测量井段的长度与相应测井曲线的垂向分辨率之比,提出了估计变点个数上限的方法;根据变点跳跃度,提出了实现并层处理的方法。将此类方法应用于某井段自然伽马测井曲线,所得的岩性划分结果与实钻录井分层相吻合,证明了该方法的有效性。

Change-point analysis, a widely applied technique in non-linear statistics, is introduced to detect lithologic boundaries from geophysical well-log data. The mean change-point model is adopted in the study. The least squares estimator is utilized to determine the number and positions of the change-points in the well-log data, which are interpreted to be lithologic boundaries. The maximum possible change-point number is pre-estimated by using the ratio of the depth interval penetrated by the log data to the minimum resolved thickness (resolution) of the well log data used, such as gamma-ray log, and commingled interpretation is performd based on the jump of change-points. The lithologic boundaries automatically identified in a case well by using this method match well with the drilling results, verifying its effectiveness.