Parameter estimation of a high-voltage transformer using pseudo-random impulse sequence perturbation

Date
2022-12
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Transformers are essential devices to the power system. Transformer applications include transmission services, monitoring of line voltages and currents, and distribution services. High-voltage test transformers are commonly used in environments when high voltages are required to test power system equipment. Due to the cost of large transformers and their importance to the power system, it is essential to monitor their health and condition to ensure nominal grid operation. Access to an equivalent-circuit model that represents the physical construction of the transformer aids in condition monitoring and identifying the location of faults. Due to the proliferation of renewable generation, there are increased harmonics in the power system. Therefore, there is a growing need for equivalent-circuit models that provide physical representation and accuracy over a wide range of frequencies. System identification can be applied to transformers to develop wideband equivalent-circuit models from measured data. In this project, a methodology for determining wideband models for transformer windings and a high-voltage test transformer is presented. The Pseudo-Random Impulse Sequence (PRIS) signal is a wideband excitation signal for system identification in the power system environment. The PRIS perturbation arrangement is applied to two cascaded transformer winding models in simulation. Successful offline perturbation of transformer winding models with the PRIS signal is verified through good correlations between the simulated and analytically derived frequency responses of the input impedance and voltage ratios. The performance of various optimisation algorithms are investigated and compared. The algorithms are applied to various approaches, namely, frequency-domain, time-domain, or modal decomposition, in combination with various cost function formulations. It is found that although the time-domain and modal decomposition approaches perform superior, frequency-domain estimation that makes use of the model’s frequency responses derived analytically is the most computationally efficient. A PRIS source is built and subsequently applied to perturb a practical 7.5kVA, 230V/100kV high-voltage test transformer. The transformer measurements show minimal noise up to 20kHz. Four equivalent-circuit model topologies are investigated to identify a suitable model and associated parameters that best reproduces the transformer measurements. It is found that a single-section low-voltage winding and a six-section high-voltage winding model topology are of a high enough order to represent the number of observed resonances up to 10kHz. The measured frequency responses and estimated frequency responses produced by the model show a good correlation, with discrepancies below 1kHz due to the lack of any frequency-dependent model parameters.
AFRIKAANS OPSOMMING: Transformators is baie belangrike toestelle in die kragstelsel, en word gebruik vir transmissie doeleindes, om lyn spannings en fout strome te meet, en ook vir distrubisie doeleindes. Hoogspannings transformators word oor die algemeen gebruik in omgewings waar ho¨espannings benodige word om kragstelsel toerusting te toets. Omdat groot transformators duur is en ’n belangrike komponent in die kragstelsel is, is dit belangrik om transformators deurgaans te monitor om nominale werking te verseker. Toegang tot die ekwivalente stroombaan model wat die fisiese bou van die transformator beskryf help om die kondisie van transformators te monitor en foute binne die transformator te vind. As gevolg van die verspreiding van hernubare opwekking, is daar verhoogde harmonieke in die kragstelsel. Daarom is daar ’n groeiende behoefte aan ekwivalente stroombaanmodelle wat nie net fisiese voorstelling verskaf nie, maar wat ook akkuraatheid oor ’n wye reeks frekwensies verskaf. Stelselidentifisering kan toegepas word om transformators te beskryf deur wyeband modelle te onwtikkel. In hierdie projek word ’n metode om akkurate modelle vir transformator windings en ’n hoogspannings toets transformator voorgestel. Die Pseudo-Lukrake Impuls Volgorde (PLIV) sein is ’n bre¨eband opwekkings sein wat vir stelsel bepalings studies gebruik kan word in die kragtselsel. Die PLIV opwekking stelsel is aangewend op twee kaskade winding transformator modelle in simulasie. Suksesvolle vanlyn perturbasie van die transformator windings met die PLIV sein is geverifieer deur goeie korrelasie tussen die gesimuleerde en analitiese frekwensie uittrees van die intree impedansie en spannings verhoudings. Die optrede van verskeie optimiserings algoritmes word ondersoek en vergelyk. Die algoritmes word toegepas op verskeie benaderings, naamlik frekwensie-domein, tyd-domein en modale ontbinding, tesame met verskeie koste funksie samestellings. Dit is bevind dat alhoewel die tyd-domein en modale ontbinding benaderings beter presteer, die analities berekende frekwensie-domein oordragfunksies van die model meer koste effektief is. ’n PLIV bron is en daarna aangewand om ’n 7.5kVA, 230V/100kV hoogspannings toets transformator te versteur. Die metings van die transformator toon minimale geraas tot 20kHz. Vier ekwivalente stroombaanmodel topologie¨e is ondersoek om die mees gepaste model en bypassende parameters te verkry. Dit is bevind dat ’n enkel-seksie lae-spanning winding en ses-seksie ho¨e-spannings winding model topologie van ’n hoog genoeg orde is om die aantal waargeneemde resonansies tot 10kHz voor te stel. Die gemete frekwensie gedrag en afgeskatte frekwensie gedrag vanaf die model toon goeie korrelasie, met verskille onder 1kHz wat volg uit die tekort aan frekwensie-afhanklike parameters.
Description
Thesis (MEng) -- Stellenbosch University, 2022.
Keywords
Parameter estimation, High voltages, Electric transformers, Perturbation (Mathematics), UCTD
Citation