The invention provides a microgrid inverter fault diagnosis method based on wavelet transformation and a probabilistic neural network. The method comprises the steps that firstly, building a fault simulation system of the microgrid inverter on simulation software, then collecting three-phase load currents under different fault types to serve as measurement signals, and extracting fault characteristic signals of the three-phase load currents through a wavelet multi-resolution analysis method; carrying out normalization processing on the extracted fault feature signal to construct a feature vector; and finally, selecting a part of the extracted fault feature vectors as training data to train the PNN model, and when a genetic algorithm is adopted to find an optimal smoothing factor of the probabilistic neural network, and when the inverter has a fault, carrying out fault positioning on a current signal according to the method, thereby realizing fault diagnosis of the microgrid inverter. The method has the advantages that the structure model is simple, determination is easy, the convergence speed is high, a Bayes optimization solution is converged, the sample adding capacity is high, retraining is not needed, the precision is high, the practicability is high, and engineering combination is easy.