Assessment of the harmonic behaviour of a utility-scale photovoltaic plant

Overett, Jonathan Stephen (2017-12)

Thesis (MScEng)--Stellenbosch University, 2017.

Thesis

ENGLISH ABSTRACT: The rapidly increasing deployment of utility-scale solar photovoltaic (PV) plants worldwide presents new technical challenges including the risk of unacceptably high harmonic voltage distortion in the utility network. PV plants affect the harmonic voltage distortion at their point of common coupling primarily by two distinct mechanisms: generation of harmonic currents due to switching action of the inverters and introduction of series and parallel resonances caused by connection of the plant to the network. The objective of this study is to investigate the harmonic behaviour of utility-scale PV plants based on the case-study of a 75 MW PV plant in the Northern Cape province of South Africa. In this work, a harmonic simulation model of a PV plant is developed in DIgSILENT PowerFactory which represents the passive network components with sufficient accuracy to study the frequency response over the harmonic frequency band from 50 to 2500 Hz. 10-minute and 3-second aggregated harmonic measurement data and event-triggered waveform data were captured over a 12-day period using an IEC 61000-4-30 Class A power quality meter installed at the plant’s 132 kV point of connection (POC). A quality assessment of harmonic measurement data shows that the accuracy of harmonic voltage and current measurements are likely to be significantly affected by quantisation and transducer error for even and higher-order harmonics. Analysis of the 3-second aggregated harmonic quantities indicates that their fluctuation during each 10-minute period is sufficiently small that the 10-minute average values adequately approximate the short-term and long-term behaviour of the plant. The harmonic behaviour of the plant is investigated by considering the relationship between harmonic voltage and current emissions and active power, reactive power and time of day. Certain harmonic voltages increase with increasing plant active power output whilst others decrease but all harmonic currents remain constant or increase with increasing active power. It was not possible to quantify the relative impact of active and reactive power variation as the plant operates in voltage control mode. Brief periods of high harmonic current distortion were noted during start up and shut down of the plant. High harmonic currents at the 23rd harmonic order correspond with a series resonance identified in the model thus validating the POC frequency sweep simulation. A comparative analysis of different published methods for assessment of the harmonic emissions contribution of the PV plant is performed. All methods show similar trends in identifying dominant harmonics and general harmonic behaviour but the magnitudes of the calculated emissions differ significantly in some instances. Where accurate frequency sweep data of the utility network and phasor measurement data are both available, the plant model enables application of the harmonic vector method to discriminate between utility and plant contributions. Simulated voltage emissions based using the manufacturer’s quoted inverter current emissions do not closely match measured emissions due to the complexity of active source interactions within the real power system. Comparison of the mean, 75th percentile and 90th percentile emissions assessments demonstrate the effectiveness of percentile assessment in eliminating outlying high distortion periods from the assessed results.

AFRIKAANSE OPSOMMING: Die vinnige toenemende implementering van grootskaalse sonkragfotovoltaïese (PV) aanlegte wêreldwyd, bied nuwe tegniese uitdagings, insluitende die risiko van onaanvaarbare hoë harmoniese spanningvervorming in die netwerk. PV aanlegte beïnvloed die harmoniese spanningvervorming by hul punt van algemene koppeling, hoofsaaklik deur twee afsonderlike meganismes: die opwekking van harmoniese strome as gevolg van die skakel aksie van die omkeerders en die inleiding van reeks en parallelle resonansies wat veroorsaak word deur die koppel van die kragstasie aan die netwerk. Die doel van hierdie studie is om die harmoniese gedrag van grootskaalse-PV aanlegte te ondersoek, gebaseer op die gevallestudie van 'n 75 MW PV kragstasie in die Noord-Kaaps provinsie van Suid-Afrika. In hierdie werk word 'n harmoniese simulasiemodel van 'n PV-aanleg ontwikkel in DIgSILENT PowerFactory, wat die passiewe netwerk komponente verteenwoordig met voldoende akkuraatheid, om die frekwensieweergawe oor die harmoniese frekwensieband van 50 tot 2500 Hz te bestudeer. 10-minute en 3-sekonde geaggregeerde harmoniese metingsdata en gebeurtenisgeaktiveerde golfvormdata is oor 'n tydperk van 12 dae opgeneem met behulp van 'n IEC 61000-4-30 klas A kragkwaliteitsmeter geïnstalleer by die substasie se 132 kV punt van verbinding (POC). 'n Kwaliteitsbeoordeling van harmoniese metingsdata toon dat die akkuraatheid van harmoniese spanning en stroommetings waarskynlik aansienlik beïnvloed sal word deur kwantiserings en oorvormer foute vir ewe en hoër-orde harmonieke. Analise van die 3-sekonde geaggregeerde harmoniese waardes dui aan dat hul fluktuasie gedurende elke 10 minute tydperk klein genoeg is, sodat die 10-minute gemiddelde waardes die korttermyn- en langtermyngedrag van die aanleg ‘n voldoende benadering is. Die harmoniese gedrag van die aanleg word ondersoek deur die verband tussen harmoniese spanning en stroom emissies en aktiewe krag, reaktiewe krag en tyd van die dag te oorweeg. Sekere harmoniese spannings verhoog met toenemende aktiewe kraguitset, terwyl ander afneem, maar alle harmoniese strome bly konstant of verhoog met toenemende aktiewe kraguitset. Dit was nie moontlik om die relatiewe impak van aktiewe en reaktiewe kragvariasie te kwantifiseer nie aangesien die aanleg in die spanningskontrolemodus funksioneer. Kort tydperke van hoë harmoniese stroomvervorming is tydens die aanvang en afsluiting van die aanleg opgemerk. Hoë harmoniese strome op die 23ste harmoniek stem ooreen met 'n reeks resonansie wat in die model geïdentifiseer is en dus valideer die POC-frekwensie sweep simulasie. 'n Vergelykende analise van verskillende gepubliseerde metodes vir die assessering van die harmoniese emissie bydrae van die PV-aanleg word uitgevoer. Alle metodes toon soortgelyke neigings in die identifisering van dominante harmonieke en algemene harmoniese gedrag, maar die groottes van die berekende emissies verskil in sommige gevalle beduidend. Waar akkurate frekwensie sweep data van die netwerk en fasor meting data albei beskikbaar is, maak die model dit moontlik vir die toepassing van die harmoniese vektor metode om te onderskei tussen netwerk en PV-aanleg bydraes. Gesimuleerde spanningsemissies gebaseer op die vervaardiger se verklaarde omkeerder-emissies, pas nie noukeurig ooreen met gemete emissies nie as gevolg van die kompleksiteit van aktiewe broninteraksies binne die werklike kragstelsel. Vergelyking van die gemiddelde 75ste persentiel- en 90ste persentielemissiesevaluasies toon die effektiwiteit van persentielassessering om die afgeleë hoë vervormingsperiode uit die geassesseerde resultate te elimineer.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/102869
This item appears in the following collections: