Research On Dynamometer Cards Soft-sensing Of Walk-beam Oil-pumping Based On Neural Network

Beam pumping unit is widely used in oilfield, the dynamometer is an effective means to analysis and diagnosis whether the normal operation of the pumping unit, and it is an important part of the daily management of the oil pumping machine. Beam pumping unit dynamometer is a diagram of the relationship between hanging point load and displacement. The working condition under the pumping oil is judged by it.Current methods of measuring the dynamometer is measured regularly, it needs to use the load sensor, so each measurement must be repeated down from machine. This approach not only complicated measuring process, operation is also low safety factor, but also affect the life of the pumping unit. In addition, because it is the regular measurement, so unable to secure the running status of oil pumping machine, resulting in oil pumping machine is not timely maintenance, the influence of oil production, a waste of energy. Therefore, real-time measurement pumping dynamometer is the fundamental purpose of further study, but also as a prerequisite for energy-efficient operation of the pumping unit.On the surface, it is seems very simple that the oil pumping machine action, actually involves a variety of simple quantitative relationship, and variations are nonlinear. Therefore, it is difficulties that real-time measurement of direct or indirect dynamometer. In order to overcome the lack of regular measurements, many scholars have proposed different methods of indicator measurement beam pumping dynamometer, but there are limitations.For the beam pumping unit of time measurement dynamometer difficult problem, put forward the research of oil pumping machine dynamometer of soft measurement based on neural network. Neural network not only has a solid theoretical foundation, and does not require a specific mathematical model, almost any nonlinear approximation. The core of the dynamometer of soft measurement is the study of the choice easier related physical quantity measurement, such as the motor power, the beam swing angle, by trained neural network to predict pumping rod load. Through the data communication means, the field data related to the data center, by the prediction of oil pumping unit rod load and displacement, constitute the oil pumping machine dynamometer.The thesis introduces three neural networks, that is, forward BP, RBF neural network and Elman feedback neural network. In this paper, the model of the structure and the working principle is described in detail and the establishment of the dynamometer of soft measurement model. Using MATLAB computer software as research tools, three kinds of neural network simulation, and the results were analyzed and compared, illustrate the rationality and feasibility of the research theory.