Research And Application Of The Fault Diagnosis Method Of Downward Barbell Rod Pump Pumping Well Based On Electric Power

At present,.more than 90%of oil wells use rod pumps to lift oil.Conventional rod pumps are simple in structure and reliable in operation,but they also have the disadvantages of poor balance performance and serious energy consumption.The lower eccentric barbell pumping unit with rod pump adds a lower eccentric body to the tail of the conventional rod pump pumping unit,and it makes the pumping unit more balanced and the motor energy consumption is smaller.The lower-biased barbell rod pump pumping unit is simple to modify,and its performance is improved compared with the conventional rod pump pumping unit,so it is widely used in oilfield production.Traditional fault diagnosis of pumping wells relies on indicator diagrams,but indicator diagram measurement is complicated and does not have real-time performance.Starting from the power information of the pumping unit’s motor,this thesis uses intelligent signal analysis and diagnosis methods as tools to study the fault diagnosis of pumping wells with rod pumps under the barbell.First,this thesis introduces the ground structure and downhole structure of the rod pumping system,expounds the pumping principle of the rod pumping system in detail,and theoretically analyzes the indicator diagram and electric power information under typical working conditions.Secondly,there is no complete electric power library for electric power information.In order to visually compare the difference of electric power information under typical working conditions,this thesis establishes a ground four-bar linkage model of a pumping well with a lower-biased barbell.This model converts the downhole indicator diagram information into ground motor power information,and it intuitively compares the electric power under typical fault conditions with the electric power under normal conditions.This thesis puts forward the time domain characteristics of electric power on this basis;Aiming at the abnormal vibration when the gearbox fails,wavelet transform is used to extract the frequency domain features;In order to ensure the relative independence between the eigenvalues,the KPCA algorithm is used to reduce the dimensionality of the features.Then,in order to ensure the balance of the data set,this thesis uses the Borderline-SMOTE algorithm to derive data from a small number of fault samples,and proves the effectiveness of the method;At the same time,BP neural network is easy to fall into the shortcomings of local optimum,so this thesis uses the improved PSO algorithm to optimize the initial parameters of the BP neural network.The simulation proves that the optimized diagnostic model has higher accuracy.Finally,with the help of the intelligent measurement and control instrument platform for pumping wells,industrial experiments are carried out on the method proposed in this thesis.The experiment has achieved satisfactory results.