Research On Fault Diagnosis Method Of Sucker Rod Pumping Wells Based On Support Vector Machine

At present,the main oil recovery technology is rod pumping technology in China.Due to the complex geographical environment and poor working conditions,the oil well failures often occur.The traditional manual well fault diagnosis method cannot have a specific understanding of the working condition of the oil well in real time.Once the fault occured,if the situation is relatively light,it will cause economic losses to the oil field,if serious,it will worn down personal safety hazards to the staff.Indicator diagram as a commonly used fault evaluation basis,contains abundant information about oilfield operation conditions.With the rapid development of computer technology,intelligent fault diagnosis technology is gradually used on the stage of fault diagnosis.In this paper,the indicator diagram is taken as the research object,and the intelligent fault diagnosis method based on the indicator diagram is applied to the fault diagnosis of oilfield for classification and recognition.First,the sucker rod pumping system is studied in detail,and the information of indicator diagram is analyzed.The indicator diagrams under different conditions were analyzed,and the characteristics of different conditions of indicator diagram are concluded,a solid foundation for the intelligent classification method was found.Secondly,the method of feature extraction is used to analyze the indicator diagram.In this paper,the Deep Belief Network is used to extract the features of the indicator diagram.We use the unsupervised greedy learning algorithm in Deep Belief Network to preprocess the indicator diagram.After we use Deep Belief Network in the fine-tuning to make the parameter adjust.Thirdly,Support Vector Machine is used to identify classification.In this paper,Support Vector Machine is used as a classification function.In dealing with the problem of nonlinear classification,SVM selects Gaussian kernel function as the high-dimensional mapping function of SVM.Aiming at the uncertainty of the parameters of SVM,we choose the Improved Artificial Fish Swarm Algorithm to optimize the parameters of SVM,so that the optimization speed has been significantly improved and the recognition accuracy is very high.Finally,we develop a host computer software system for real-time fault detection in oilfield.In order to detect the oil field in real time,this software system uses Visual Basic 6.0 programming language to realize the design of monitoring interface,and realizes the functions of each module of the upper computer software system.The upper computer software runs well in oilfield operation,realizes the real-time fault diagnosis of sucker rod pumping wells,meets the anticipated requirements and meets the needs of actual production.