Abstract: In order to detect short duration power quality disturbance correctly，a new classification method of short duration power quality disturbance based on K-L transformation and multiclass support vector machine is proposed in this paper.Original feature space is constructed using discrete wavelet transformation.Pattern recognition feature space is distilled using K-L transformation.A multiclass support vector machine suitable for short duration power quality disturbance is designed.The simulation results show that the classification correctness can be improved using K-L transformation.The classification result using multiclass support vector machine is better than BP neural classifier.