Gaussian Filters for P Class PMUs: a Performance Comparison of Alternative FIR Design Procedures

Protection-oriented (P Class) Phasor Measurement Units (PMUs) are required to estimate the synchrophasor, the frequency and the Rate of Change of Frequency (ROCOF) of ac voltage or current waveforms with high accuracy and low latency to detect possible anomalous events in power systems promptly and effectively. The classic architecture of many commercial PMUs relies on the direct frequency Down-Conversion and Filtering (DCF) of the collected input waveform. However, the adopted low-pass filters are generally optimized for harmonic disturbance rejection, while no much attention is usually devoted to overshoot minimization, which is instead of crucial importance when sudden phase or amplitude steps occur. Recalling that in principle the Gaussian filters exhibit zero-overshoot in the case of step changes, in this paper two alternative Finite Impulse Response (FIR) approximations of the Gaussian filter (one based on the windowing method and the other on the cascaded boxcar filters, respectively) are proposed and compared. Several simulation results (obtained in the P Class testing conditions of the IEEE/IEC Standard 60255-118-1:2018) confirm that both filters provide better results than those obtained with the classic triangular impulse response filter suggested in the Annex D of the Standard. Also, the results in the case of step changes exhibit almost zero overshoot and the same response times. However, due to the different frequency response in the stopband, the Gaussian filter approximation based on the cascade of boxcar filters provide slightly more accurate results than in the case when the windowing method is used, even though the latter provides a better approximation of the Gaussian impulse response.