Research on Production Condition Identification Method of Pumping Well Based on an Improved Pattern Recognition Algorithm

In the process of oil and gas production, indicator diagram is an important basis for production management. Diagnosing the working condition of pumping unit based on indicator diagram can quickly grasp the working state of oil well, analyze and judge whether the production parameters of oil well are reasonable. The pattern recognition technology of indicator diagram generally includes two stages: feature extraction and classification. Classification methods mainly use a few common machine learning methods such as artificial neural network and SVM. In recent years, feature extraction methods mainly include geometric feature method, grid feature and moment invariant feature methods. Due to the limitations of these extraction methods, most of them can only extract some primary feature quantities. However, using these primary feature quantities to recognize the indicator diagram often has the technical problems of low recognition efficiency and low recognition accuracy. In this paper, an improved indicator diagram pattern recognition method based on binary gradient direction histogram feature is proposed. Compared with the previous feature extraction methods, this method has good robustness, fast calculation speed and high recognition accuracy.