Reservoir Permeability Prediction Based on Machine Learning

This paper discusses the challenges and solutions of carbonate reservoir permeability prediction. It is difficult to predict the permeability of carbonate reservoir because of its complex pore structure and heterogeneity. Although NMR logging is useful for characterizing pore structure, its high cost limits its wide application. Therefore, in this paper, machine learning technology and petrophysical logging data are used to compare the permeability prediction models. By constructing and testing permeability prediction models based on artificial neural network (ANN), decision tree (DT) and support vector machine (SVM), the advantages and limitations of various methods are analyzed, which provides new tools and methods for oil and gas exploration and development.