Effective voltage flicker detection approach based on a new modified S-transform algorithm

Higher standards have been placed on the flexibility and accuracy of characteristic detection methods of voltage flicker characteristics due to sporadic flicker disturbances. S-transform as a tool of signal analysis has been limited in practical application because of the fixed shape of its Gaussian window function. Some existing modified S-transform algorithms have demonstrated increased adaptability and flexibility by an additional parameter in Gaussian window function. However, the additional parameter is selected by empirical data in some modification, which have exposed errors brought by human factors. In this paper, a new improvement is achieved by introducing the measurement criteria of time-frequency concentration which can be used to optimize the additional parameter of Gaussian window function. The core of this scheme is to obtain an optimal parameter through maximizing the value of measurement criteria of time-frequency concentration. The algorithm is tested on a set of synthetic signals and a voltage flicker disturbance signal. The results show it can demonstrate much more detection accuracy and the ability to resist the cross term interferences than other modification.