贝叶斯统计识别法在录井油气评价中的应用

Application of Bayesian statistical pattern recognition in mud log oil-gas evaluation

综合录井技术广泛应用于油气勘探活动中,而油气资源评价是勘探活动中最主要的工作之一。针对原有传统油气评价方法的不足,探索用模式识别技术进行统计分析。通过对录井特征参数的选择,采用了窗函数法和近邻法求其特征分布,并以此来构造概率密度分布函数,并应用贝叶斯(Bayes)判别决策方法训练分类的有关参数,确定了基于线性窗函数的贝叶斯方法和基于独立近邻原则的贝叶斯方法。最后分别采用传统图版解释法和统计模式识别法进行了对比分析,使其油气评价的符合率得到上升,为油气评价方法提供了一种新的思路。

Mud log technique has been used widely in oil-gas exploration, and oil-gas resource evaluation is one of most main work of them. Aimed at the shortage of traditional method, pattern recognition technology is explored to make statistical analysis. By choosing mud log feature parameters, a method to get distribution function is presented by using window function and neighborhood, and furthermore constructs probability density function. Bayes classification decision is applied to train related parameter, and two synthesis recognition method are proposed, which are Bayesian method based on liner window function and Bayesian method based on independent neighborhood principle. Finally,real data using both traditional graphical method and statistical pattern recognition method are analysed. The result identifies that the ratio of actually is climbed. It provides a new path for oil-gas evaluation method.