Research And Implementation Of Soft Sensing Method Of Pumping Unit's Indicator Diagram Based On Electric Parameters

The sucker rod pumping wells are working at thousands of meters underground,and they are under the action of periodic alternating load for a long time,it is easy to produce many kinds of faults.It is necessary to make timely diagnosis of the working conditions of the oil wells.Detection of indicator diagram is the basis of oil well condition diagnosis.In view of the shortcomings of the existing dynamometer detection technology,this paper focuses on the soft-sensing method of obtaining indicator diagram with electrical parameters of motor.This paper established a hybrid model of indicator diagram soft-sensing based on SVR,and designed a indicator diagram soft-sensing hardware platform,which is applied in industrial field.Firstly,this dissertation analyzed the principle and model of sucker rod pumping system.According to the working principle of beam pumping unit,the kinematics analysis of four-bar linkage and crank mechanism was carried out,and the functional relationship between suspension displacement,acceleration and rotation angle was established.Through the analysis of the force and torque transfer relationship of crankshaft,and the mathematical function relationship between the suspension load and the instantaneous input power of the motor was established,which provides a theoretical basis for the establishment of the indicator diagram soft sensing model.Secondly,this dissertation established a soft sensing hybrid model of indicator diagram based on support vector regression.At first,through the establishment and simulation of the mechanism model,this paper analyzed and corrected the non-convergence of the fitted dynamometer at the upper and lower dead points.Then,aiming at the errors of fitting dynamometer diagram in the upper and lower dead points and fluctuation of dynamometer diagram,using support vector regression to compensate the mechanism model.Combined with the mechanism model,established a hybrid indicator diagram soft-sensing model.Simulating and comparing the result of hybrid model with the indicator diagram of measured dynamometer diagram,data-driven model and mechanism model,the result shows the error rate is less than 5%.At last,the dissertation analyzed the factors affects the accuracy of indicator diagram.Finally,according to the actual demand and production situation of oil field production,this research designed a indicator diagram soft-sensing hardware platform on the basis of model research.The platform achieved the real-time analysis and monitoring of the indicator diagram,electrical parameters and motor operation status of pumping units.At the same time,a simple and easy-to-operate interface is developed to make the field operation more intuitive and easy to use.At last,the platform is applied to the oil well production field to carry out experiments,and has achieved high accuracy,proved that it has high practical value.