Power quality disturbance identification and classification are important for power quality management, for which a new method is proposed to deal with power quality disturbance identification and classification based on PSOELM (particle swarm optimizationextreme learning machine). Decompose the disturbance signals with wavelet for ten layers and extract the layers of energy difference which can effectively distinguish the difference between disturbance signals, the average of energy difference and the standard deviation of energy difference as feature vectors, in addition, the root mean of disturbance signals and normal signals is calculated as a supplement in order to reduce the dimension of the importing vectors. It is proposed that ELM training error is used as the fitness function of PSO to optimize the hidden layer neuron number to enhance the speed of classification and also maintain high classification accuracy. Simulation results show that this method can accurately and effectively identify seven common disturbance types and have a higher classification speed comparing to the traditional BP neural network.