Classification of Reservoir Fracture Development Level by Convolution Neural Network Algorithm

Abstract

Identifying and classifying fractures is an important task in the study of fractured oil and gas reservoirs. The most common solution is to identify it by artificial interpretation or synthetically probability methods, and to classify them according to the degree of fracture development. In order to improve the accuracy and reduce the man-made or computational errors, this study introduces the convolutional neural network (CNN) algorithm, one of the deep learning algorithms, to distinguish the degree of fracture development while constructing a new model which can automatically identify cracks and determine the category of fractured reservoirs in the meantime. Firstly, the logging curves with strong sensitivity to fractures are selected as the input data of convolution neural network, and the crack category is quantified as the output label of the network. A CNN model which is suitable for the classification of cracks is designed, whose parameters is continuously optimized through a small batch gradient descent method in the training stage. Then the trained convolutional neural network is applied to process the logging data of an oil field. The comparison of the result of crack classification by convolutional neural network with that by the traditional BP neural network indicates that the unique convolutional weight sharing structure of convolutional neural networks can extract the most effective features and greatly improve the accuracy of the fracture classification in dealing with complex nonlinear problems such as the classification of fractured reservoirs.