In order to accurately identify all kinds of voltage sag sources in a power grid and avoid the difficulty of feature extraction experienced by other methods, this paper proposes an identification method of voltage sag source based on distance discriminant analysis from the perspective of minimum distance. The characteristics of waveform change of voltage sag RMS are analyzed using the sag segmentation method, and six populations corresponding to the types of voltage sag source are constructed based on a coarse-grained RMS waveform. The multi population Mahalanobis distance discriminant analysis method is adopted, and the training samples are used for learning, and the corresponding discriminant functions and discriminant criteria are established to distinguish the samples to be judged, so as to realize the identification of voltage sag sources. The proposed method is validated and compared by simulation modeling. The results show that the proposed method has high recognition accuracy and performance, low cross error rate, good noise robustness and meets the practical application requirements. This work is supported by National Natural Science Foundation of China (No. 51807126).