Aiming at lack in transformer fault diagnosis based on DGA ,support vector machine was proposed based on particle swarm optimization method of transformer fault diagnosis.The support vector machine classifier of transformer fault diagnosis model was established, this thesis uses the particle swarm optimization algorithm and libSVM toolbox on matlab platform to train the support vector machine classifier. Finally, it uses well trained the support vector machine to forecast 110 kv star transformer substation fault condition. The results show that based on particle swarm optimization support vector machine in transformer fault diagnosis is consistent with the actual results.This method can enhance the accuracy of transformer fault diagnosis.