Aiming at the problem of power quality detection and classification, a method for classifying and identifying the faults of power quality based on KPCA (Kernel principal component analysis) and FDA (Fisher Discriminant Analysis) is proposed. Using KPCA to extract the characteristics of the energy index, and the high-dimensional information of the index is deeply excavated. According to FDA, the extracted principal components are classified with high precision, and the training results are adjusted by the training array and the test array. Finally, the central eigenvector of the six types of power quality disturbance is determined, and the detected power quality data are classified. In accordance with the experimental results, the KPCA-FDA is used to classify the six types of power quality, which are more precise than PCA and KPCA in the correctness of various power quality fault classification.