The invention discloses a transformer fault diagnosis method based on feature information quantization and weighted KNN, comprising the following steps: S1, dividing sample data into a training set and a test set; S2, inputting the training set, and performing preprocessing on the sample data; S3, based on principal component analysis (PCA) and grey relational analysis (GRA), performing quantization on fault feature information; S4, introducing a particle swarm optimization algorithm for optimizing a weighted KNN categorization algorithm, according to a true fault category, training a sample in a standardized fault feature matrix, and obtaining a power transformer fault diagnosis model, thus categorization on a power transformer fault is realized; and S5, inputting the test set into the power transformer fault diagnosis model, and obtaining a diagnosis result, thus diagnosis on the power transformer fault is realized. The transformer fault diagnosis method disclosed by the invention solves the problems that processing efficiency is low, model training is difficult and limitation exists in the prior art.