A Study On Diagnosis Of Sucker-rod Pumping System Based On SVM

In recent years, rod pumping oil dominates the main position in the oil industry. Oilfield environment is harsh, complex well conditions, oil well equipment is easy to damage, serious impact on the oil recovery efficiency and economic benefits. So grasp the oil well conditions is very important to achieve the automatic monitoring and scientific management of production systems.Although there are many diagnostic methods, but these methods makes similar indicator diagrams recognition rate very low. This article rod pumping oil system diagnostic method is the support vector machines, neural networks, wells working parameters combination. Rod pumping system fault diagnosis rate can be improved by using a high degree of recognition ability of this method on a small sample of events.In-depth analysis of the rod pumping oil system works and dynamometer forming principle based on the sucker rod pumping system conditions, and highlights of the typical conditions dynamometer graphics features. Extract measured indicator diagram of moment invariants((?),i=1,....7), well pump efficiency(ηv), on the stroke the measured load with the theoretical load deviation (△F1), deviation of the measured load and theoretical load of the down stroke (△F2) represents the characteristic parameters of the well conditions, the establishment of a rod pumping system typical conditions characteristic parameter sample library; and support vector machine classifier rod pumping system diagnostic method based on statistical learning theory and support vector machine theory. By constructing a support vector machine rod pumping system diagnostic experiments show that:the input sample pretreatment, the choice of the kernel function and parameter (candg) optimization will affect the end result. In order to verify the advantages of this method, support vector machines and neural networks (BP and RBF) for the comparative analysis of conditions in the same wells. The results prove that the support vector machine diagnosis capabilities and higher recognition accuracy in small samples and similar well conditions.Through the results showed that:Integrated features [(φ,i=1,....7),η,△F1,△F2)]combined support vector machine rod pumping system fault diagnosis method can not only solve the problem of similar indicator diagram which is difficult to judge but also has a strong ability toidentify well conditions.