Research On Parameter Identification And Time Compensation Of Electromagnetic Actuator Of Rail Transit Circuit Breaker

Rail transit circuit breaker is the equipment of disconnecting and connecting the main circuit of vehicles and high voltage of the contact line.Its intelligent operation has become a hotspot of research.Synchronous control technology as one of the contents of intelligent operation of the circuit breaker is of great significance in practical application.To meet requirement of synchronous control for precision and reduce the dispersion of action time,action time of the circuit breaker operation mechanism should be compensated.At present,time compensation of the synchronous control system of most circuit breakers only takes into account control of effects of voltage and temperature on the action time but does not take into account the specific circuit parameters.In addition,existing time compensation algorithms are batch learning algorithms which cannot realize compensation for the change of opening and closing time arising from wear of the contact and aging of the mechanical structure.In this thesis,the electromagnetic actuator of the rail transit circuit breaker is taken as the research object,and a kind of Bayesian network action time compensation method in combination of online parameter identification is put forward.Firstly,the operating principle and dynamic properties of the electromagnetic actuator of the rail transit circuit breaker are analyzed,and influential factors of the opening and closing time of the electromagnetic actuator are determined.Circuit parameters which affect the opening and closing time of the electromagnetic actuator are identified with the RLS.Simulation analysis is carried out for the identification precision and the rate of convergence of parameter identification,which verifies the reliability of the online parameter identification algorithm.On the basis of parameter identification,the opening and closing time of the electromagnetic actuator is predicted with the Bayesian network.The Bayesian network action time prediction model is established in Matlab.Then comparison between the model prediction results and the prediction results of BP neural network model which only takes into account the control voltage and temperature are carried out.It is found through comparison of the mean absolute error and the root mean square error that the prediction values of the Bayesian network model are closer to the real values,indicating that the precision of action time prediction of the Bayesian network model in combination with online parameter identification is superior to that of the BP neural network model.The incremental learning characteristic of Bayesian network is used to dynamically adjust the Bayesian network action time prediction model according to the new data.,the prediction error of the model has always been within 0.25 ms,which verifies that the Bayesian network prediction model can realize compensation for opening and closing time change arising from wear of the contact.