The open circuit fault of the switch tube may be caused by the failure of the driving board, poor line contact or over-current burning. At this time, it will not cause serious consequences immediately, but there will be problems such as the increase of harmonic of inverter output waveform, the increase of voltage and current stress of switch tube, and the imbalance of bus capacitor voltage. If the fault is not found and eliminated in time, it may cause further damage to the inverter.
Previously, scholars at home and abroad have done a lot of fruitful research on inverter fault diagnosis, which can be divided into three types of fault diagnosis methods based on current, voltage and intelligent algorithm.
A normalization method based on Fourier transform is proposed, which realizes the open circuit fault diagnosis of two-level inverter and solves the problem that the traditional average current method is easy to be misdiagnosed when the load suddenly changes. Combining the normalized load current with the duration of the load current in the zero-crossing region, the open circuit diagnosis of multi-switch tubes in two-level inverter is realized, which is more reliable than the traditional normalized load current method.
The combination of the average value of load current and the average value of absolute value of load current in a current cycle realizes the open circuit fault diagnosis of multi-switch tubes, which effectively improves the robustness of fault diagnosis. Neutral point clamp, NPC) three-level inverter is used to accurately locate the single tube open circuit fault. Taking the polarity of load voltage and current at the output side of inverter as fault characteristics, the open circuit fault diagnosis of multi-switch tube of NPC three-level inverter is realized. It has the advantages of fast diagnosis speed and high accuracy, but it needs to add additional voltage sensors.
A fault diagnosis method based on support vector machine (SVM) is proposed, which uses the two-dimensional relative centroid characteristics of fault current trajectory to classify faults, and realizes the accurate location of single-tube open fault of two-level inverter, but the calculation of wavelet decomposition and sample training is large.
The output voltage of the inverter is a fault feature. Neural network is used to classify faults, and the open circuit fault of NPC three-level inverter is diagnosed quickly and accurately, but more voltage sensors are needed, which requires a lot of calculation.
To sum up, the fault diagnosis method based on voltage increases the complexity of the circuit; The fault diagnosis method based on intelligent algorithm will greatly increase the operation and processing time of fault diagnosis; The fault diagnosis method based on current does not need to add additional sensors, nor does it need to carry out complex transformation on the sampled signal, but only needs to extract the fault current information for a period of time to realize accurate fault diagnosis.