Contents I. Introduction
Computational devices used to control power facilities fall into two groups: those for slow (seconds-long)processes and those for fast (milliseconds-long)processes. The latter include power-system protection (PSP)and fault localization (FL), whereby faults include short circuits (SC)and open-phase faults. In case of FL, SC can be eliminated as a part of the automatic re-activation cycle; or the line can be switched off for a longer period until a repair team handles the fault. Given that SC have to be switched off within two or three frequency periods, PSP should operate within one or two periods. Thus, emergency control algorithms have to meet very stringent requirements. In such cases, we need algorithms that non-industrial-frequency currents and voltages can't interfere with. Modern digital PSP algorithms are triggered after analog or digital filtering is complete [1]–[3]. Filtering undergoes the transient-response stage, which is time-consuming. This is especially true when it comes to filtering such SC currents that contain both periodic and non-periodic components. Dev-proposed PSP provide for a protection delay for at least one period [1], [4]. The situation is exacerbated by the possible saturation of current-transformer magnetic cores. Saturation-induced errors and other errors can be reduced by means of least squares [5], [6] or by discrete Fourier transformation [7]–[11]. Cosine filters are often used in microprocessor protection [10]–[15]. Cosine filters are practically insensitive to non-periodic components. However, these methods take long to implement. This is why it is important to develop algorithms capable of isolating the sinusoidal components within half a period or less. This is what our research is about.
