This paper presents a new approach for open-circuit fault diagnosis using the park’s vector technique and K-means clustering algorithm in a three-phase voltage source inverter. Firstly, The method performs Park transformation on the system output signal, extracts the centroid of the fault pattern in a single cycle and the Fourier amplitude of the dq plane angle as the feature vector, then uses the K-means clustering algorithm to perform feature classification, and finally, simulation verification in three-phase inverter circuit. The accuracy of inverter fault diagnosis is effectively proved. The accuracy of fault detection is reached 99.99%, and the accuracy of fault diagnosis is over 99.65%. Simulation results show that the method is feasible and effective.