| Compared with the two-level inverter,the three-level inverter has been widely used in the field of medium and high voltage and high power due to its advantages such as low output harmonic content,high voltage utilization rate and small switching loss.But the number of three level inverter switching devices,failure rate also rose,either switch fails,will cause severe fluctuation of output current distortion,output voltage,the neutral point potential imbalance and other issues,such as not accurate and effective to discrimination of fault type,missed the best processing time,eventually lead to the entire circuit to stop working,and endanger the life property safety,Therefore,timely fault diagnosis and fault-tolerant control is the key to ensure the safe and stable operation of the three-level inverter.In this paper,a Neutral Point Clamped(NPC)three-level inverter is taken as the research object to analyze and model the inverter fault types,and the fault diagnosis is carried out by wavelet packet transform and Support Vector Machine(SVM).Finally,the space vector redundancy strategy and topology reconstruction strategy are used for fault-tolerant control.The work is divided into three parts.Firstly,the topological structure,working principle and fault causes of the three-level inverter are analyzed.On this basis,the fault types are divided and coded uniformly.Based on Matlab/Simulink platform structures,inverter open circuit fault simulation model,the simulation of different fault happens after three phase current waveform diagram,analyze the time-frequency characteristics of inverter output current distortion,using the single tube and double tube fault current waveform change characteristic,choose three phase current signals as an original fault fault feature extraction.Secondly,a fault diagnosis method based on the combination of wavelet packet transform and support vector machine(SVM)is proposed to solve the problems of complex calculation and low accuracy of open-circuit fault diagnosis.The original three-phase current signals with their own time-frequency characteristics were decomposed into eight frequency bands by wavelet packet transform,and the 24 D fault eigenvectors were constructed by wavelet packet energy reconstruction,which were used as the input of support vector machine.Particle Swarm Optimization(PSO)was used to optimize the parameters of SVM,and radial basis kernel function was used to obtain multi-classification SVM model from the training data set.The randomly selected test set was input into the trained model to test its classification accuracy.The simulation results show that this method can diagnose the open-circuit fault of switching devices and has higher fault identification accuracy.Finally,two fault-tolerant control strategies are proposed to solve the problem of fault-tolerant topology and control of NPC three-level inverter.According to the different characteristics of the space vectors before and after the fault,the single-phase bridge arm faults are divided into two categories: when the Ⅰ type of fault occurs,the space vector redundancy control strategy is adopted,and the space vector lost after the fault is replaced by the redundant vector;When Ⅱ type faults occur,the fault-tolerant control strategy based on topology reconstruction is adopted to reconstruct the original topology structure into the fault-tolerant topology structure of three-phase eight-switch inverter.At the same time,considering the midpoint voltage balance problem of the DC side of the inverter,the zero-sequence voltage component is injected to suppress the fluctuation of midpoint potential.The simulation results show that the above two fault-tolerant control strategies can provide constant amplitude voltage and current,and effectively suppress the fluctuation of DC bus voltage. |