Modeling And Fault Diagnosis Of Circuit Breakers With Spring Operated Mechanism

The scale of power system increases fast and its structure becomes more and more complex.Therefore,much attention has been paid to the safety and reliability of electrical equipments.The circuit breaker,an important unit in the field of power system,has been a hot research topic for a long time.Surveys have shown that the contact wear,non-synchronous closing and partial discharge of circuit breakers will seriously affect the safety and stability of the power system and cause huge economic losses.To solve these problems,it is common practice to monitor circuit breakers on line,identify the features of the measured data,evaluate their operating status,and diagnose faults if there are any.A large database is required to diagnose and analyze the faults of a circuit breaker,however,the amount of data obtained only by collecting field data or simulating on the physical experimental platform is not enough.Also,it is a pretty hard problem to obtain all the features of all types of equipments,and the difficulties and the costs of the implementation of field test are also unacceptable.Thus,this thesis presents an idea to build the coil current model,energy storage motor current model and arc model of circuit breakers in MATLAB and Simulink,and then generate a large number of feature signals.Afterwards,the thesis proposes two optimization algorithms-genetic algorithm(GA)and stochastic optimization algorithm based on GA,to optimize the model parameters,in order to make sure the signals obtained from MATLAB and Simulink are basically the same as the actual signals.Therefore,a variety of operating data can be obtained by modifying the parameters of the models,which provides reasonable and sufficient data for subsequent feature extraction and fault diagnosis.For the measured data,this thesis also involves a signal processing method based on mathematical morphology,which is used to filter the characteristic signals of the circuit breakers.This thesis proposes a fault diagnosis method for circuit breakers based on fast template matching.Firstly,the K-means clustering algorithm is used to cluster the relevant data of the circuit breakers.After forming a standard library,fast template matching is used to diagnose faults.As the waveforms corresponding to measured data of each type of fault are different,this method can diagnose faults effectively,and it also has a small amount of computation.In addition,this thesis proposes a fault diagnosis method of circuit breakers combined with deep belief network and Softmax classifier.The application of deep belief network can not only extract the high-level information of the characteristic signals,but also reduce dimension of the data,which avoids the negative impact of the large dimension on the classification results.The method based on fast template matching is applied to the scenario that has few kinds and the difference between each kind is evident,while the method based on deep belief network is effective in the big data scenario.Analyzing the state of the circuit breaker comprehensively through the above methods provides a good foundation for the safe operation of the circuit breaker.