The invention provides a method for monitoring transformer's fault state based on Spiking neural network. The method comprises the following steps: using the characteristic gas method to determine the number of neural elements in the input layer and the output layer of the Spiking neural network and to determine the input sample data; determining the number of neural elements in the hidden layer of the Spiking neural network; normalizing the input sample data; coding the input sample data; and inputting the input sample data to the Spiking neural network to have the data trained; outputting the trained data as the monitoring result for the transformer's fault state. In view of the monitoring problem with the transformer's fault state, the method of the invention uses an oil-filled power transformer as a study object. And the Spiking neural network adopted by the method employs the coding mode of accurate pulse time, which is closer to the real biological neural system and enables it to have a stronger computing ability than other neural networks so as to effectively identify the type of fault, reduce erroneous determination and the erroneous determination rate and provide accuracy for diagnosis.