Abstract: To solve the problem that insulated gate bipolar transistor wiring fault of photovoltaic inverters are difficult to be found and are easily to be ignored, and the existing diagnostic methods are slow and have low recognition rate, a feature extraction strategy based on voltage mean is proposed, and the kernel extreme learning machine(KELM)is optimized using improved loin swarm optimization(ILSO)to achieve fault diagnosis of IGBTs. Firstly, the effect of the Concordia transformation of the three-phase voltage in each state of the inverter is analyzed to obtain a two-dimensional vector that clearly characterizes the fault, and the separability of the wiring fault is verified by a two-dimensional scatter plot. Secondly, the parameters of KELM are optimized by the LSO of Sine chaotic mapping to establish the diagnostic model. Finally, the Z-source inverter is taken as an example for verification. The results show that the proposed method can extract two-dimensional fault features containing rich fault information for IGBT wiring faults of inverters, which greatly simplifies the process of feature extraction. Moreover, the proposed method is suitable for single and double device fault types. Compared with other methods, the proposed method is faster and has higher fault recognition rate.

Key words: inverter, wiring fault, average voltage, lion swarm optimization, KELM, fault diagnosis