Identification of power quality disturbances based on adaptive generalized S-transform and probabilistic neural networks

This paper presents a new generalized S-transform with a self-adaptive window width factor based on the Fourier spectrum of disturbances signals. The adaptive generalized S-transform (AGST) has better time-frequency resolution for disturbances recognition than traditional S-transform. Firstly, the main frequency component is calculated from the Fast Fourier transform spectrum of disturbance signals. The types of disturbances are rough separated into subsets. Secondly, the value of window width factor is adaptive changed with the requirement of time and frequency resolution for disturbances recognition. Finally, seven features are extracted from the results of AGST and the new classifier based on probabilistic neural networks is constructed for recognizing ten types of disturbances include two types of complex disturbances. The simulation experiments show that the simple new method has high classification accuracy and anti-noise ability.