The invention discloses a priori probability-based multivariate power grid fault type judgment method, which comprises the following steps of: 1, obtaining historical fault data from a multivariate power grid database, and carrying out feature extraction on the historical fault data; 2, calculating the prior probability of feature classification output of each type of historical fault data extracted in the step 1; 3, performing feature extraction on the test data sample, and performing calculation according to the prior probability obtained in the step 2 to obtain fault classification probability output corresponding to the test data sample feature and the historical fault data feature; and 4, performing weighting processing on the probability output of the test data sample characteristicsin each type of faults, and outputting the fault classification with the maximum weighting result as a final judgment fault classification result. According to the judgment method, the dimension of data operation can be reduced, and the judgment efficiency of the system can be improved.