Research on Measurement and Prediction of Well Blowout Fluid Flow Rate based on Machine Vision

Abstract

A machine vision-based method for measuring and predicting well blowout fluid flow rate is proposed to solve the problems of limited quantitative identification of well blowout fluid flow rate, inability to install traditional measurement equipment in the field, and inability to accurately measure and predict blowout fluid flow rate. This method establishes an image processing model based on ORB-BF-RANSAC for blowout sequence images, feature extraction and feature identification are performed on the blowout images from consecutive frames in order to obtain the coordinates of the feature points within the well blowout images. The pixel displacement of the feature point pairs is then calculated. Subsequently, the true displacement of the feature points is calculated using the pinhole imaging principle. Finally, the true displacement is divided by the number of frames to calculate the blowout fluid flow rate. Compared to the speed of mass flow meter calibration, the method achieves a measurement accuracy of 91%. Based on the above method, a well blowout fluid flow rate prediction method is designed by introducing Spatial Attention Mechanism (SAM) to improve the Graph Convolutional Networks (GCN). The prediction model proposed in this paper is more effective compared with other prediction methods, and the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) are 12.39%, 18.72% and 9.16%.