The invention relates to a microgrid fault diagnosis method for optimizing an extreme learning machine based on a whale algorithm. The method comprises the steps of: S1, building a microgrid grid-connected operation simulation model comprising a wind driven generator, a photovoltaic cell and a storage battery, and collecting three-phase fault voltage signals at two ends of a line; S2, selecting adb6 wavelet as a wavelet basis, decomposing and reconstructing the three-phase fault voltage signals containing the phase A, the phase B and the phase C obtained by simulation according to a wavelet packet analysis related formula, calculating the energy entropies of the three-phase fault voltage signals to obtain a feature vector X= [x1, x2,..., x24] T containing 24 wavelet packet energy entropies, and taking the feature vector as a data sample; and S3, utilizing a whale algorithm WOA to optimize an input weight and a hidden layer threshold of an extreme learning machine ELM to establish a WOA-ELM fault diagnosis model, and substituting the data sample obtained in the S2 into the WOA-ELM model to carry out training and verification. A BP neural network, an RBF neural network and the ELM are utilized to establish the diagnosis model, the diagnosis model and the WOA-ELM model are subjected to comparative analysis, and the effectiveness and reliability of the WOA-ELM model are verified.