Application Of Deep Learning In The Dynamometer Card Recognition

Petroleum resources are still the most important resources today,and the oil production is related to the lifeline of the country’s economy.At present,over 90% of the oil production equipment in our country is the rod-pumping unit.Rod-pumping unit works in a very harsh underground environment,subject to sand,steam,wax,corrosion and other factors,often wax,sand,pump leakage,sucker rod fracture and other problems,interrupt oil production.The collection and analysis of the dynamometer card of the rod-pumping unit is an effective measure and the main measure to detect,prevent and solve various faults in the oil production process.However,the recognition of the dynamometer card mainly depends on the labor and is influenced by the experience of the supervisory staff.Therefore,using new computer technology to automatically and accurately recognize the dynamometer card has been rodpumping system working conditions monitoring research focus.Deep learning has injected new vitality into the research of classification and recognition of the dynamometer card,and the deep convolutional neural network has been applied in image classification.This paper deeply studies the principle and structure of deep convolutional neural network models,and uses them as classifiers to carry out the automatic classification and recognition of dynamometer card.According to the experiment’s results,the model that is more suitable for the automatic classification and recognition of the dynamometer card is selected.Based on the chosen model,new model which has better performance than the origin model is put forward.Experiments show that the new model can better meet the practical requirements of automatic classification and recognition of dynamometer card.Then,this paper uses the new proposed model as the classifier designed and implemented a complete condition monitoring and reverse control system.The system can collect real-time dynamometer card and judge the working condition of rod-pumping unit,then based on the recognition results the system can reverse control the rod-pumping unit.For oilfield automation provides a new solution.