A partial discharge pattern recognition algorithm based on statistical characteristics parameters and probability neural network is designed for the transformers. Two-dimensional diagrams are constructed from the phase resolved partial discharge. Then, the statistical characteristics parameters are extracted from the two-dimensional diagrams. The statistical characteristics parameters are regarded as the input of probability neural network. The proposed algorithm is compared to the typical algorithms of partial discharge type recognition on corona discharge, surface discharge and air-gap discharge in experiment. It is indicated by the experimental results that the accuracy recognition rate of the proposed algorithm is higher, and the time cost is smaller. This work is supported by National Natural Science Foundation of China (No. 51477070).