The Research On Fault Diagnosis Method Of PV Three Level Inverter

The rapid development of the photovoltaic industry has led to the widespread application of photovoltaic microgrids.In the photovoltaic microgrid,the DC output from the PV array must be converted into AC power through the inverter to meet the user's requirements.Compared with the traditional two-level inverters,multi-level inverters have been widely used due to advantages such as serial voltage equalization,small switching loss,low output harmonic content,and high working efficiency.However,the number of power components in a multi-level inverter is large,so the probability of failure will increase.Any failure of any component will cause the circuit to operate abnormally,resulting in incalculable economic losses.Therefore,when a fault occurs in a multi-level inverter,it is particularly important to quickly and accurately locate the faulty component.Taking the most widely used midpoint-clamped three-level inverter as the research object,the fault feature extraction and diagnosis methods are studied in depth.The main research content is as follows:(1)Set up a mid-clamped three-level inverter circuit model in MATLAB,and as the experimental platform for follow-up research and verification.By removing the IGBT pulse signal to the open circuit fault of the power element,and the output waveforms of the three-phase currents under various fault types are obtained.(2)Because there are many types of faults in three-level inverters,it is difficult to diagnose them directly.According to the characteristics of higher three-phase current waveform identification in single-phase fault and low three-phase current waveform identification in two-phase fault,this paper divides the fault into two categories: typical and atypical,and carries out diagnosis separately.(3)According to the characteristics of the corresponding power waveforms of the positive and negative half-wave parts of the output current in the normal and fault of each phase,the extreme values of the superposed power waveform corresponding to the positive and negative half-wave parts of the output current of each phase are proposed as fault feature quantities.This method can not only reduce the dimension of the feature quantity,but also have the function of automatically normalizing the fault data.(4)By analyzing the extreme characteristics of the superimposed power waveform under typical faults,a diagnosis method based on interval coding for typical faults is proposed.The test results show that this method has the characteristics of fast diagnosis,highaccuracy and strong anti-noise ability.(5)The particle swarm optimization(PSO)was used to optimize the weights and thresholds of BP neural network,and applies it to diagnose atypical faults of three-level inverters.Compared to a single BP classifier,the PSO-BP classifier has higher diagnostic accuracy and better robustness.(6)Using particle swarm optimization(PSO)to optimize the penalty parameter and the kernel function parameter of the support vector machine(SVM),and applies it to diagnose atypical faults of three-level inverters.The PSO-SVM classifier has better diagnostic performance than a single SVM classifier.