Research On Fault Diagnosis Method Of Pumping Well Based On Support Vector Machine

Nowadays,rod pumping technology is mainly used in oil production in China.However,most of the oil fields are located in the complicated geographical environment,so the production status of oil wells can only be monitored by the way of intermittent inspection of oil wells by workers in the oil field,so there is no way to achieve real-time monitoring of the working status of oil wells.If the oil well fails to find out and take measures in time,the light one will damage the pumping unit,resulting in the decrease of pumping efficiency,and the heavy one may lead to the leakage of crude oil,causing significant losses.Because the indicator diagram contains the information of the working condition of the oil well,this paper takes the indicator diagram as the research object,and then analyzes the indicator diagram to classify,identify and diagnose the fault of the rod pumping well.Firstly,the working principle of rod pumping well is analyzed,then the formation reason of indicator diagram is simply introduced,and the characteristics of indicator diagram under several typical working conditions are introduced in detail,which paves the way for fault diagnosis later.Secondly,the method of feature extraction of indicator diagram is studied.In this paper,we first use Hu moment invariants to extract the features of the indicator diagram,then use Freeman chain code to extract the features of the indicator diagram,and finally use the data fusion method to fuse the feature vectors obtained by the two methods serially.After that,SVM is used to identify,classify and diagnose faults.The kernel function of SVM is Gauss kernel function.Then the improved artificial bee colony algorithm is used to optimize the parameters of SVM,which improves the speed of optimization and has high recognition accuracy.Finally,the Intelligent monitor of pumping well is developed.It realizes the real-time analysis and diagnosis of the working condition of the pumping well.When the intelligent monitor is installed,the monitor will collect the relevant data,then extract the characteristics of the indicator diagram,and then diagnose the fault of the extracted characteristics.The final diagnosis results can be displayed in the interface of human-computer interaction.In the middle process,there is no need for the staff to monitor on the spot,which realizes the high performance.The process of efficiency and convenience has achieved the purpose of intelligent diagnosis of oilfield conditions,which is conducive to oilfield production.