主成分分析在低孔低渗储层孔隙度预测中的应用研究  

Application of Principal Component Analysis in Porosity Prediction of Low Porosity and Low Permeability Reservoir

作　　者：张雨辰 赵军龙 孙婧 崔文洁 陈家鑫 金利睿 ZHANG Yuchen;ZHAO Junlong;SUN Jing;CUI Wenjie;CHEN Jiaxin;JIN Lirui(Xi′an Shiyou University,Xi′an 710065,China;Shaanxi Key Laboratory of Petroleum Accumulation Geology,Xi′an 710065,China)

机构地区：[1]西安石油大学地球科学与工程学院,陕西西安710065 [2]陕西省油气成藏地质学重点实验室,陕西西安710065

出　　处：《河北地质大学学报》2025年第2期21-31,共11页Journal of Hebei Geo University

基　　金：国家自然科学基金面上项目(42172164)。

摘　　要：为了提高低孔低渗储层孔隙度预测的精度,更好地服务于低孔低渗储层测井评价,通过梳理主成分分析的原理性质,结合机器学习常用方法和集成学习组合策略,提出利用选择性集成学习的思想构建基于主成分分析的孔隙度预测模型。首先对归一化后的测井曲线数据进行主成分分析,然后将提取出的主成分作为BP神经网络、RF(随机森林)、XGBoost(极致梯度提升树)、岭回归四种机器学习模型的输入属性,最终利用优选算法按照比重构建集成模型对孔隙度进行预测。研究表明,该模型对低孔低渗储层的孔隙度预测值与实际值的相关系数R^(2)为0.948,预测精度较高,优于单一机器学习预测模型;解决了传统低孔低渗储层孔隙度预测方法精度不足、泛化能力较差等问题,为后续低孔低渗储层的测井综合评价奠定了基础。In order to improve the accuracy of porosity prediction in low porosity and low permeability reservoirs and better serve the logging evaluation of low porosity and low permeability reservoirs,by combing the principle and properties of principal component analysis,combined with the common methods of machine learning and ensemble learning combination strategies,the idea of selective ensemble learning is proposed to construct a porosity prediction model based on principal component analysis.Firstly,the principal component analysis of the normalized logging curve data is carried out,and then the extracted principal components are used as the input attributes of four machine learning models:BP neural network,RF(random forest),XGBoost(extreme gradient boosting tree)and ridge regression.Finally,the optimization algorithm constructs an integrated model according to the specific gravity to predict the porosity.The research shows that the correlation coefficient R^(2)between the predicted value and the actual value of the porosity of the low porosity and low permeability reservoir is 0.948,and the prediction accuracy is higher,which is better than the single machine learning prediction model.It solves the problems of insufficient accuracy and poor generalization ability of traditional porosity prediction methods for low porosity and low permeability reservoirs,and lays a foundation for subsequent logging comprehensive evaluation of low porosity and low permeability reservoirs.

关 键 词：主成分分析法 孔隙度预测 集成学习 低孔低渗储层 

分 类 号：P618.13[天文地球—矿床学]