THE APPLICATION OF A KPCA-AVM MODEL TO RESERVOIR IDENTIFICATION
PANG He-qing1, KUANG Jian-chao1,2, WANG Zhong1, LIU Hai-song1, CAI Zuo-hua3, HUANG Yao-zong4
1. College of Energy Resources, Chengdu University of Technology, Chengdu 610059, China;
2. College of Management Science, Chengdu University of Technology, Chengdu 610059, China;
3. Guiyang Department of Exploration and Development Institute, Southwest Petroleum Branch, Sinopec, Guiyang 550004, China;
4. No.29 Party of No.4 Gudong Oil Factory, Shengli Oilfield, Sinopec, Dongying 257237, China
Abstract：It is more difficult to predict the low porosity and low permeability tight reservoir than to predict the regular reservoir.The authors therefore tentatively applied kernel principal component analysis and support vector machine,called KPCA-SVM model,to solve this problem.Through the polynomial kernel function of the KPCA,the model can obtain the nonlinear feature extraction.Then the Gaussian kernel function in the SVM is chosen to perform optimization again.Finally,reservoir identification is implemented in the SVM.As the model incorporates the advantages of kernel function,principal component analysis and support vector classification,it can better solve the problem of nonlinear small sample,eliminate the noise of the data and reduce the dimension without missing valid information.In addition,it can achieve the prediction function quickly and accurately.The model was employed to predict the reservoir in x856 well block,which belongs to Xu2 member gas reservoir of the Xinchang gas field.The prediction results show the superiority of this model,which can be used as an optional method in tight reservoir prediction.
庞河清, 匡建超, 王众, 刘海松, 蔡左花, 黄耀综. 核主成分分析与支持向量机模型在储层识别中的应用[J]. 物探与化探, 2012, 36(6): 1001-1005，1013.
PANG He-qing, KUANG Jian-chao, WANG Zhong, LIU Hai-song, CAI Zuo-hua, HUANG Yao-zong. THE APPLICATION OF A KPCA-AVM MODEL TO RESERVOIR IDENTIFICATION. Geophysical and Geochemical Exploration, 2012, 36(6): 1001-1005，1013.
