doktorska disertacija
Jalen Štremfelj (Author), Dušan Agrež (Mentor)

Abstract

Tematika doktorske disertacije posega na področje merjenja in vrednotenja parametrov harmoničnih signalov v frekvenčnem prostoru. Meritve tovrstnih električnih parametrov so pomembne za delovanje elektroenergetskih sistemov tako iz inženirskega kot tudi ekonomskega vidika. V delu so predstavljene metode za ne-parametrično računanje električnih signalov z neznano frekvenco, amplitudo, ter faznim zamikom. Zaradi čedalje večje uporabe nelinearnih elementov, merjeni signali v omrežju pogosto vsebujejo tudi harmonske komponente, katerih velikost včasih ni zanemarljiva. Ker so merjeni signali po večini sinusoidne oblike in vsebujejo več komponent, je smiselno analizo izvajati v frekvenčnem prostoru. Tak pristop omogoča boljši vpogled v razmere v omrežju, hkrati pa lahko vnaša določen merilni pogrešek zaradi končnega časa merjenja. Z daljšanjem časa merjenja velikost pogreška sicer upada, problem pa nastane pri merjenju kratkotrajnih prehodnih pojavov dolžine nekaj period. V takšnih situacijah je potrebno uvesti metode za glajenje merjenih signalov. Največji del pogreška pri analizi v frekvenčnem prostoru se pojavi zaradi nestacionarnosti signalov in nekoherentnega vzorčenja. Pri slednjem čas trajanja meritve ni enak mnogokratniku frekvence merjenega signala, zato se po izvedbi diskretne Fourierjeve transformacije v frekvenčnem spektru pojavijo določene harmonske komponente, ki sicer v merjenem signalu niso prisotne. Dodatno se merjenemu signalu spreminjata amplituda in frekvenca, kar še dodatno otežuje izvedbo koherentnega vzorčenja. V disertaciji so opisani 3 novi algoritmi za zmanjšanje ocenjenega sistematičnega pogreška oziroma vrednotenje merilne negotovosti tipa B z uporabo diskretne Fourierjeve interpolacije (DFT) v kombinaciji z Rife-Vincentovimi okenskimi funkcijami 1. reda (RV-1). Okenske funkcije so uporabne za zmanjševanje pogreška odtekanja v frekvenčnem spektru, ki je posledica nekoherence v časovnem prostoru. Močna okna izraziteje odpravljajo pogrešek iztekanja, ob tem pa zaradi širokega glavnega grebena slabše vplivajo na ločljivost blizu ležečih komponent. Interpolacija merjenih oziroma izračunanih parametrov se vrši v frekvenčnem prostoru s seštevanjem dveh, treh ali več največjih DFT koeficientov, s čimer se zmanjša sistematični pogrešek iztekanja. Z višanjem števila uporabljenih koeficientov se zmanjšuje pogrešek iztekanja, vendar se ob tem povečuje število računskih korakov in zahtevnost izračuna.

Keywords

moč;frekvenca;frekvenčni prostor;interpolacija;pogrešek vrednotenja;merilna negotovost;doktorati;

Data

Language: Slovenian
Year of publishing:
Typology: 2.08 - Doctoral Dissertation
Organization: UL FE - Faculty of Electrical Engineering
Publisher: [J. Štremfelj]
UDC: 621.317(043.3)
COBISS: 87429891 Link will open in a new window
Views: 211
Downloads: 32
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Measurement of harmonic signal parameters in electric power systems in the frequency domain
Secondary abstract: The topic of the doctoral dissertation is in the field of measuring and evaluating the parameters of harmonic signals in frequency domain. Measurements of such electrical parameters are important for operation of power systems from both an engineering and economic point of view. The paper presents methods for non-parametric calculation of electrical signals with unknown frequency, amplitude, and phase shift. Due to the increasing use of nonlinear elements, the measured signals in the network often contain harmonic components, the amount of which is sometimes not negligible. Since the measured signals are mostly sinusoidal in shape and contain several components, it makes sense to perform the analysis in frequency space. Such an approach provides a better insight into the situation in the network, and at the same time can introduce a certain measurement error due to the final measurement time. As the measurement time increases, the magnitude of the error decreases, but the problem arises when measuring short-term transients of a period of several periods. In such situations, it is necessary to introduce methods for smoothing the measured signals. The largest part of the error in frequency space analysis occurs due to nonstationarity of signals and incoherent sampling. In the latter case, the duration of the measurement is not equal to the multiple of the measured signal frequency, so after performing a discrete Fourier transform, certain harmonic components appear in the frequency spectrum, which are otherwise not present in the measured signal. In addition, the amplitude and frequency of the measured signal change, which makes it even more difficult to perform coherent sampling. The dissertation describes 3 new algorithms for reducing the estimated systematic error or estimating the measurement uncertainty of type B using discrete Fourier interpolation (DFT) in combination with Rife-Vincent 1st order window functions (RV-1). Window functions are useful for reducing the outflow error in the frequency spectrum due to incoherence in time space. Strong windows eliminate the leakage error more significantly, but due to the wide main ridge, they have a worse effect on the resolution of the close-by lying components. The interpolation of the measured or calculated parameters is performed in the frequency domain by summing the two, three or more largest DFT coefficients, thus reducing the systematic leakage error. Increasing the number of coefficients used reduces the leakage error, but at the same time increases the number of calculation steps and the complexity of the calculation.
Secondary keywords: power;frequency;frequency domain;interpolation;evaluation error;measurement uncertainty;
Type (COBISS): Doctoral dissertation
Study programme: 1000319
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za elektrotehniko
Pages: 162 str., [22] str. pril.
ID: 14047815