Oil Reservoir Classification Based on Convolutional Neural Network

摘要

According to the existing oil reservoir data, the efficient and the accurate classification of the oil reservoir is a key factor for the petrochemical enterprises to improve the efficiency of resources utilization. However, the chromatogram data of the oil reservoir has characteristics of high dimensions, complexity and noise. Therefore, a classification model based on the convolutional neural network(CNN) is proposed to learn automatically features from the sequence data of the oil reservoir and avoid the complex feature engineering. The synthetic minority over-sampling technique(SMOTE) algorithm is used to dispose unbalanced categories samples to reduce the rate of the misclassification. Finally, the proposed method is applied in J16 oil dataset taken from an oil field of China petrochemical industry. The result shows that the accurate classification of the oil reservoir can achieve about 80.8%.