基于主成分分析的极限学习机方法开展杂卤石识别

川中地区岩性复杂,杂卤石周围广泛发育硬石膏、岩盐和其它非钾盐矿物,常规测井解释方法很难准确识别杂卤石。以极限学习机理论、主成分分析方法和测井解释为基础,把主成分分析得到的影响杂卤石识别的主变量(测井曲线)作为输入,建立极限学习机(PCA-ELM)杂卤石的精确识别模型,对比川中地区录井结果,PCA-ELM的杂卤石识别正确率达到90.74%;再以不同岩石在测井曲线上的响应特征为基础,建立杂卤石分类识别模型,分类识别正确率达到89.56%。与常规测井解释方法相比,具有速度快、操作简单、准确率高等特点。结果表明,在四川盆地钾盐勘探中PCA-ELM法是一种值得推广使用的方法。

Based on the Principal Component Analysis and the Extreme Learning Machine method to identify the polyhalite

The lithology is complex in middle Sichuan region, and anhydrite, rock salt and other non-potash minerals have widely developed around the polyhalite so that it is difficult to accurately identify the polyhalite by conventional logging interpretation method. Based on the theory of Extreme Learning Machine, Principal Component Analysis method and logging interpretation methods, this paper creates extreme learning machine model with the input of main variables which have an influence on polyhalite recognition and discriminates the polyhalite reservoirs in middle Sichuan region. Compared with logging data, the accuracy rate of the discrimination results reaches 90.54%. The new model shows the accuracy rate reaches 89.56% to classify the polyhalite reservoirs. Compared with the conventional logging interpretation method, it has the characteristics of fast speed, simple operation and high accuracy. This study demonstrates that PCAELM is worthy of further study and widely use in potassium exploration of Sichuan Basin.