LDA和KNN算法在随钻测井火成岩分类的应用

Application of LDA and KNN Algorithms to Igneous Rock Classification in Logging While Drilling

渤中34-9油田在开发过程中广泛钻遇古近系火成岩,由于火成岩岩性多样、成分复杂导致常规测井解释图版识别岩性精度较差,而在随钻过程中准确识别火成岩岩性是工程上规避憋、卡、漏等风险的重要前提。通过将机器学习算法线性判别分析(LDA)与KNN算法运用于油田开发过程中的随钻测井数据处理与分析,实现了随钻过程中准确、高效识别火成岩岩性的目的。进一步将线性判别分析的降维结果代替原始测井曲线作为K最近邻分类器的输入,实现两种算法的有机融合,并对油田5口开发井建立的测井数据集进行机器学习,火成岩岩性分类准确率高于90%,证明了该方法的适用性。通过引入机器学习方法为常规录、测井数据的处理与解释提供了新方法,多方法的结合也为油田勘探作业过程中的分类提供借鉴。

Paleogene igneous rocks are widely encountered in the development process of Bozhong 34-9 oilfield.Due to the diverse lithology and complex composition of igneous rocks,the accuracy of lithology identification by conventional logging interpretation plate is poor.However,accurate identification of igneous rocks during drlling is an important prerequisite for avoiding the risks of holding back,jamming and leakage.By applying machine learning algorithms LDA and KNN to the processing and analysis of logging data while drlling in the process of oilfield development,the purpose of identifying igneous lithology accurately and efficiently in the process of drilling is realized.Furthermore,the dimension reduction result of LDA was used as the input of KNN classifier instead of the original log curve to realize the organic integration of the two algorithms.The machine learning was carried out on the log data set established in 5 development wells in the oilfield,and the accuracy of igneous rock lithology classification was higher than 90%,which proved the applicability of the method.The introduction of machine learning provides a new method for the processing and interpretation of conventional logging and logging data,and the combination of multiple methods also provides a reference for the classification problems in the process of oilfield exploration.