Intelligent Diagnosis Technology For Remote Monitoring Pumping Well Conditions

It is a major issue urgently to be resolved in current pump production system to grasp working conditions of pumping well and to achieve remote monitoring and scientific management. With the development of automation in oilfields, some oilfields have achieved remote monitoring pumping well and therefore enormous real-time production data of oil well can be obtained; however, traditional working condition diagnosis methods can hardly meet field application demand, so intelligent diagnosis technology of pumping wells condition is studied on the basis traditional condition diagnosis. According to the mechanical analysis on sucker rod unit, the paper establishes pump indicator diagram calculation model and discusses convergence conditions for finite difference solution and calculation approach for damping coefficient through adopting finite difference method to solve the model. Ten fault causes, such as sucker rod parting, and the main shape characteristics of the corresponding pump indicator diagrams are analyzed. Through studying the theory of invariant moment and the shape characteristics of pump indicator diagram, seven invariant moment, pump efficiency, producing fluid level, load increase in upstroke, load decrease in down stroke are extracted as the characteristic parameters of pumping well condition and sample database of characteristic parameters for typical conditions of pumping well is established based on the pretreatment of the above eleven characteristic parameters. The basic structure of BP neural network, BP algorithm modification, network generalization ability, network integration and network parameter settings are studied and BP artificial neural network widely used in fault diagnosis field is used as intelligent diagnosis for pumping well conditions. In order to improve the accuracy of diagnostic system for pumping well conditions, four sub-networks are integrated to design the diagnosis model in this paper. During designing the sub-networks, Empirical formula method, partial hide nodes rejection method, dispersion coefficient method and gray correlation analysis method are applied to calculate the number of hide nodes. Based on theoretical study, intelligent diagnosis software for remote monitoring pumping well conditions is designed, trained and debugged using Visual Basic 6.0. The software is tested using real-time oil well data and results show that the method is a feasible intelligent diagnosis method and research on intelligent diagnosis technology for remote monitoring pumping well conditions has showed certain theoretical significance and practical value.