Abstract: The authors conducted model test by using three typical electrode mode of partial discharge, Through analysis and calculation the partial discharge signal characteristics were extracted, and the corresponding partial discharge signals feature library was established. The extracted feature amount of partial discharge signals was used as the fuzzy ART network's input in order to identify the type of partial discharge, and compare with the recognition results of the BP neural network. The results of experiment showed that fuzzy ART neural network was effective for pattern recognition of transformer partial discharge.