基于PCA和KNN的碳酸盐岩沉积相测井自动识别

Automatic Identification of Carbonate Sedimentary Facies Based on PCA and KNN Using Logs

碳酸盐岩油藏地质建模和储量计算中人工判别未取心井沉积相的标准不统一。以中东Y油田碳酸盐岩为例,采用主成分分析(PCA),选取累计方差贡献率大于90%的主成分代表输入的测井信息,通过正演去掉油层对电阻率的影响,运用均值滤波去齿化和众数滤波确定边界;采用岩心刻度方法,以岩心井的测井数据和沉积亚相建立学习样本,对未取心井使用K最邻近分类算法(KNN)进行沉积亚相分类预测。结果表明,KNN对沉积亚相的预测精度达到90%以上。与传统的人工神经网络(ANN)和自组织映射(SOM)预测结果比较后认为,该技术有效解决了学习样本量大、类域交叉多的难题,且运行速度快,分类结果可靠、稳定。

In geological modeling and reserves calculation of carbonate reservoir, there is no uniform criterion for identification of sedimentary facies. The purpose of this paper is to study a set of automatic identification technology for carbonate sedimentary facies. Taking Y oilfield in the Middle East for example, principal component analysis （PCA） has first been adopted. The principal components are selected according to more than 90% cumulative variance contribution. At the same time, influence of oil on resistivity is removed by forward modeling, high frequency is cleared up by median filter and boundary is gained by mode filter. With core calibration, learning samples are established based on well logging data and sedimentary subfacies of core. Then K-nearest neighbor algorithm （KNN） is used to predict sedimentary subfacies of uncored wells. The result shows prediction accuracy of sedimentary subfacies is above 90%. By comparing other methods such as Artificial Neural Network （ANN） and Self Organizing Map （SOM）, the technology is more suitable for a large number of learning samples and much classification overlap, and the prediction result is more reliable and stable.