Development of an Oil-Well Downhole Behavior Classification Method Based on Pressure Parameters

Automated processing of downhole engineering parameters and intelligent classification of working conditions play an important role in improving production efficiency and promoting the production process of oil fields. However, in the offline interpretation process of long-term downhole pressure parameters, the huge amount of data storage and uncertain downhole noise and interference can seriously affect and delay the interpretation of engineering parameters. At the same time, inaccurate and incomplete collection of downhole parameter feature extraction methods, improper selection, and optimization of downhole behavior classification mechanisms, etc., can seriously restrict the rapid and accurate extraction of pressure features, making the correct classification of downhole working conditions face greater challenges. In this paper, wavelet, time-domain, and gradient features of the downhole pressure parameters were extracted, and three methods including, neural network, support vector machine and K-nearest neighbor method, were used to classify the working conditions. By comparing the processing results under the interaction of different types of features and different behavior classification mechanisms, a classification method for downhole working conditions based on pressure parameters was optimized. The results show that the application of the optimal behavior classification method based on the optimal multi-feature has both small deviation in pressure parameters and high accuracy in working condition classification. This method provides a feasible data analysis method for downhole parameters.