Research On Intelligent Control System For Fracturing And Flowback In Tight Gas Wells

At present,there is little research on the real-time prediction algorithm of the diameter of the injector nozzle in the automatic control system of fracturing fluid return ground.The empirical method has a large prediction error and cannot be controlled in real time.The PID algorithm cannot accurately simulate the complex nonlinear function relationship of downhole multiphase flow,resulting in inaccurate prediction of nozzle diameter and affecting the effect of the control system.Therefore,the artificial intelligence method based on deep learning neural network to predict the diameter of the oil nozzle is studied,which has the advantages of high accuracy and good real-time.The contents of this article are as follows:(1)The basic model of fracturing fluid backflow system,including the downhole multiphase flow mathematical model,was established,which laid a theoretical foundation for the feasibility of deep learning neural network to predict the diameter of the oil nozzle.(2)A new enhanced batch normalized convolutional neural network(AU-BN-CNN)structure is established,which is based on the ordinary convolutional neural network model,which improves it and adds a batch normalization layer to make it adapt to the occasion of fracturing reflow.The forward and reverse propagation algorithms of AU-BN-CNN neural network parameter learning are studied.(3)Collect fracturing backflow site data,such as pressure,temperature,flow,viscosity,etc.,combine it with the numerical solution of the downhole multiphase flow mathematical model,and preprocess and transform these data to form AU-BN-CNN neural network training samples and test sample sets.(4)The AU-BN-CNN network was simulated and experimented.Based on Matlab software,the AU-BN-CNN neural network model was established and simulated.The results show that the prediction accuracy of the AU-BN-CNN network is 99.6%,the prediction accuracy of Le Net-5network is 86.9%,the prediction accuracy of AlexNet network is 93.1%,and the prediction accuracy of CNN network is 95.7%,so the prediction accuracy of AU-BN-CNN network is the highest,which verifies the superiority of AU-BN-CNN network.Finally,an experimental platform was built to verify the method,and the automatic adjustment of the oil nozzle was observed by entering different pressure,flow,temperature and viscosity.Through 12 hours of observation,the loopback fracturing fluid in the pipeline did not sand out,which verified the effect of the AU-BN-CN network.Through the study of this paper,hydraulic fracturing technology in the oil and gas industry can be promoted,and this method can also be applied to similar occasions.