Application of BP Neural Network in Efficiency Prediction of Oilfield Mechanized Mining System

• Published in: Journal of failure analysis and prevention 2022, Vol.22 (2), p.658-665
• Main Authors: Zhou, Yong, Wang, Yang, Wen, Xi, Han, Chuanjun
• Format: Article
• Language: English
• Subjects: Back propagation networks; Characterization and Evaluation of Materials; Chemistry and Materials Science; Classical Mechanics; Corrosion and Coatings; Economic impact; Economic models; Efficiency; Energy conservation; Independent variables; Materials Science; Mathematical models; Neural networks; Oil field equipment; Oil fields; Oil wells; Parameters; Prediction models; Quality Control; Reliability; Safety and Risk; Solid Mechanics; Technical Article > -Peer-Reviewed; Tribology
• ISSN: 1547-7029; 1728-5674; 1864-1245
• DOI: 10.1007/s11668-022-01360-6

Setting production parameters that can match well with the current situation of oil well production is one of the direct means to save energy and improve the efficiency of oil field mechanical production system. However, in some oil wells and production areas, because of bad environment and aging equipment, there are few production parameters that can be adjusted, and the parameters need to be adjusted frequently. To reduce the economic cost caused by multiple parameter adjustments, an efficiency prediction model based on the neural network is proposed for this kind of oil wells, which can effectively predict the system efficiency after parameter adjustment. The independent variables of the ANN prediction model were determined by the grey correlation degree, and then, the BP ANN prediction model between the efficiency of the ANN was established by selecting the oil well data of the same test site and permeability degree and produced fluid parameters close to each other as samples, and the model was verified by field experiments. The experimental results show that the BP neural network can effectively predict the efficiency of the machine mining system. The maximum relative error between the actual efficiency and the neural network prediction efficiency after the parameter adjustment is 9.32%. It can provide reference basis and effect prediction for the actual parameter adjustment in the field and realize the purpose of energy saving and efficiency increase.