The invention discloses a power distribution network low-voltage prediction method based on multi-source evidence fusion. The method comprises the following steps: to begin with, selecting multiple indexes for reflecting power distribution network low-voltage generation possibility; collecting voltage detection values corresponding to different index values under various indexes; obtaining prototype eigenvectors of low-voltage and non-low-voltage propositions supported by various indexes by utilizing a fuzzy logic neuron clustering network learning method; then, serving various types of indexvalues corresponding to a distribution network to be predicted as evidences for low-voltage predication thereof, establishing a basic trust distribution function for various kinds of evidences by utilizing an optimum clustering method; and finally, carrying out fusion on the multiple low-voltage predication evidences through a D-S evidence synthesis rule, and according to the fusion result, predicating power distribution network low-voltage generation possibility. The method can predicate power distribution network low-voltage phenomenon, can enable the prediction result to be more objective and reliable through multi-source information fusion, and provides scientific basis for technical improvement project for reasonable arrangement and treatment of the low-voltage phenomenon.