A method of electric power system transient signals recognize based on similarity of wavelet time entropy matrixes

Transient protection technology is undergoing rapid development, the main problem its practicality faced is how to recognize fault and non-fault transient. The wavelet time entropy, defined on the basis of signal wavelet analysis, can represent the time-frequency features of transient signal in many scales at the same time. Also it is robust for noise and transient amplitude. In power system, the time-frequency features of different transient signals can't be the same. This should cause the difference between wavelet time-frequency entropy distributions under multi-scale. Based on this characteristic and wavelet time entropy matrix, a classification method using these matrixes' similarity is presented in this paper. This similarity is like the one in digital picture process. The method presents standard time-entropy matrixes of all transients, and then calculates the matrixes of testing sample and entropy similarity. Make sure that the classification can reach the maximal value of similarity as possible. The transient simulations of power system fault, normal breaker switching, capacitor switching and lighting disturbance show the method is effectively in fault recognition.