Fault Diagnose Research Of Cascaded Multilevel Inverter Based On BP Neutral Network

Cascaded multilevel converters have the advantages of modularization and easy expansion.However,the cascaded multilevel topology is mainly used in high-voltage and high-power occasionsand their overload capacity is limited,and the switching device works in the high-frequency switching state,with large loss,serious heating,and the maximum probability of failure.The actual operation shows that most of the failures of power electronic equipment arc the damage of the switching device.In this paper,fault information is presented only through a single current sensor.The fault diagnosis of cascaded inverters is realized by using the method of locating fault types.The validity and correctness of the proposed method are verified by simulation in order to improve the reliability of the system.Firstly,the working principle and topology of cascaded multilevel inverters are introduced,and then the control and modulation strategies are elaborated.The mathematical model of cascaded multilevel inverters is analyzed.The working principle of cascaded multilevel inverters based on diode clamped three-level topology and the current flow paths in different states are analyzed.The voltage closed loop and control strategy are analyzed.On this basis,the carrier phase disposition(PD),the carrier phase opposition disposition(POD),the alternative phase opposition disposition(APOD),and the carrier phase shift(CPS)are analyzed.The modulation strategies are then determined,which are suitable for cascaded multilevel inverters.Due to the faults of cascaded multi-level inverters are complex and the number of switches is large,this paper begins with the analysis of current flow paths under different faults from a single module,and then analyses different fault characteristics.In the case of different faults of switches,the current THD on DC bus is very suitable.In view of this characteristic,this paper proposes a method to locate the fault type only by extracting the fault information of a single current sensor,that is,to locate and detect the fault through the difference of current THD of DC bus.After observing different sub-harmonics and phases,the DC side current is finally selected.The magnitude and phase of the 0,2,3,4 th harmonic after Fourier decomposition are used as fault characteristics.The data samples and the appropriate algorithm are obtained by simulation,and a two-layer back propagation neural network structure is established.The eigenvectors are input into the back propagationneural network.Through simulation,the accurate fault diagnosis results are obtained,and the accuracy and validity of the back propagation neural network are verified.Finally,on the MATLAB/Simulink simulation platform,the cascaded multilevel inverters are modeled and simulated under different switching tube faults.The simulation results verify the feasibility and validity of the proposed fault diagnosis scheme for single current sensor.