ANALYSIS OF ALGHORYTHMS FOR PMU AT VARIOUS STATIC AND DYNAMIC SIGNALS

V elektroenergetskem omrežju se zaradi njegove kompleksnosti in povečanega vključevanja alternativnih virov sočasno pojavljajo nestabilnosti, ki zmanjšujejo eno izmed njegovih najpomembnejših lastnosti, zanesljivost. Z obstoječo merilno tehniko in regulacijo je bilo težko slediti novim izzivom, zato so se razvili PMU-merilniki (angl. »phasor measurement units«). PMU merilniki so naprave, ki v realnem času beležijo ključne parametre na elektroenergetskem omrežju (amplitudo, frekvenco, fazo in spremembo frekvence) in to stanje nato v realnem času posredujejo nadzornikom omrežja. Jedro vsakega takega merilnika je mikroprocesor z vgrajenim algoritmom, ki iz vzorčenega signala v določenem časovnem oknu izračuna parametre omrežja, pri izračunu pa napake ne smejo presegati vrednosti določene v standardu IEEE Standard for Synchrophasor Measurements for Power System (IEEE Std. C37.118.1-2011). V svojem diplomskem delu sem uporabil simulacijsko platformo implementirano v okolju Matlab in v njej testiral delovanje osmih algoritmov. Pri testiranju sem uporabil različne statične signale (linearna deviacija amplitude, faze ali frekvence ter harmonska popačenja) in dinamične signale (amplitudna stopnica, amplitudna rampa ter frekvenčna stopnica in frekvenčna rampa), kot jih bodisi zahteva IEEE Std. C37.118.1-2011 standard ali so sicer pogosto prisotni v elektroenergetskem omrežju. Kot neprimerna sta se izkazala algoritma FFT (angl. »fast Fourier transform«) in 3PSF (angl. »three parameter sine fit«), saj predpostavljata, da se frekvenca ne spreminja in je vedno enaka 50 Hz. Algoritma 4PSF (angl. »four parameter sine fit«) in MHFE (angl. »multi harmonic frequency estimation«) sta manj primerna v primeru harmonskih popačenj, medtem ko je delovanje iDFT3 (angl. »three point interpolated DFT with Hann window«), SLCA (angl. »spectrum leakage correction algorithm«) in PSFEi (angl. »interpolated phase sensitive frequency estimation«) algoritmov slabše v primeru frekvenčne rampe. Glede na vse izvedene simulacije se je kot najboljši izkazal PSFE (angl. »phase sensitive frequency estimation«) algoritem.Power grids have become less stable and reliable in the recent years due to the grids complexity as well as increasing number of renewable sources. New challenges were difficult to follow with existing measurement technology and regulations, therefore the phasor measurements units (PMUs) have been developed. The PMUs are devices that record crucial grids parameters like amplitude, frequency, phase and rate of change of frequency in real time and then send the data to grid operators. The core of each PMU is a microprocessor with embedded algorithm that estimates grids parameters from the signal sampled within the certain observation time interval. Several different algorithms have been developed and investigated in the recent years and their performance differs significantly. However, the errors of the estimation process should never exceed values specified in IEEE Standard for Synchrophasor Measurements for Power System (IEEE Std C37.118.1-2011) otherwise the algorithms should not be implemented in the PMU. In my thesis, a simulation platform in Matlab environment was used to test how eight algorithms response to different static signals (i.e. linear deviation of amplitude/phase/frequency or harmonically distorted signals) as well as dynamic signals (amplitude/frequency step/ramp). These test signals are either required by the IEEE Std. C37.118.1-2011 or the signals are readily presented on the power grid. The simulations showed that the fast Fourier transform and three parameter sine fit algorithms are not appropriate since a constant nominal frequency 50 Hz is always presumed. The four parameter sine fit and multi harmonic frequency estimation showed poor performance when the signal was harmonically distorted while the three point interpolated DFT with Hann window, spectrum leakage correction algorithm, interpolated phase sensitive frequency estimation algorithms showed poor performance when a frequency ramp was applied. On the contrary the phase sensitive frequency estimation algorithm showed the best overall performance