Research on the Method of Convolutional Neural Network in Oil Well Drawing

The indicator diagram is a method for judging the type of failure of the pumping unit system. Since many oilfields still recognize the collected dynamometer by manual analysis, there is an error in manual identification. Due to the complicated working environment of the pumping unit, the pumping system will encounter various problems and cannot accurately identify the type of fault in time. In this paper, the fault indicator diagram is taken as the research object, and the convolutional neural network is the theoretical basis. With reference to the design idea of the more mature convolutional network model, the basic network is optimized. Then, the Bagging algorithm that improved the voting mechanism was improved, and the accuracy rate of 0.7% was improved. On this basis, the recall rate was 4.56% higher. Meet the needs of the actual production environment, with good predictive effects and practical performance.