Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Modelling and short-term forecasting of high wind speed events at operational wind farms
Thesis (PhD)--Stellenbosch University, 2019.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA |
| Published: |
2019
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614114391523328 |
|---|---|
| access_status_str | Open Access |
| author | Groch, Matthew |
| author2 | Vermeulen, H. J. |
| author_browse | Groch, Matthew Vermeulen, H. J. |
| author_facet | Vermeulen, H. J. Groch, Matthew |
| author_sort | Groch, Matthew |
| collection | Thesis |
| description | Thesis (PhD)--Stellenbosch University, 2019. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/107184 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA |
| last_indexed | 2026-06-10T12:46:53.692Z |
| license_str | Not specified — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2019 |
| publishDateRange | 2019 |
| publishDateSort | 2019 |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/107184 Modelling and short-term forecasting of high wind speed events at operational wind farms Groch, Matthew Vermeulen, H. J. Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering. Winds -- Speed Wind farms Wind energy conversion systems Artificial neural network Thesis (PhD)--Stellenbosch University, 2019. ENGLISH ABSTRACT: The increasing penetration of wind energy has prompted a reform of the prototypical operational practices of conventional power systems, especially systems dominated by thermal generation sources. The variable nature of wind energy generation requires that further investments be made into more flexible plant with faster start-up capabilities to safeguard against potential shortfalls in generation. These rapid response services are dispatched by the System Operator in reaction to large power ramps to ensure that energy balance is maintained. Rapid response services, however, typically assumes the form of storage or gas turbines, which are procured at a high cost. In the context of wind power ramping phenomena, High Wind Speed Shutdown (HWSS), potentially, represents the most severe risk to power system stability. It is clear from the available literature, that HWSS has not been extensively investigated to date. Although the need for forecasting and quantification of the impacts of HWSS feature strongly in the available literature, no models have thus far been formulated to describe this phenomenon, and no event-based forecasting models have been proposed in response to this research question. This dissertation targets two major aspects of HWSS, namely the modelling, quantification, and comparison of the relative risk of HWSS events, and the short-term operational forecasting of HWSS events. It is evident from the literature that the development of a dedicated HWSS forecasting model will assist in the management and mitigation of the short-term risk associated with HWSS events. The development of site-specific models with which to quantify and compare temporal risk will, furthermore, aid in the siting of wind farms in regions with a low susceptibility for HWSS events. Two novel event-based forecasting techniques are proposed for the short-term forecasting of HWSS events, namely an Artificial Neural Network (ANN) model approach, and a hybrid model using an original statistical downscaling methodology. Both of the proposed model topologies utilise an ensemble wind speed forecast derived using the Weather Research and Forecasting (WRF) model, as well as additional environmental variables such as wind direction and temperature. The results prove that both models demonstrate good accuracy for the forecasting of localised high wind speed events which occur at the micro-scale level, which is in line with HWSS events. A technique is proposed for site characterisation and comparison of HWSS events for wind farm site planning. The proposed technique utilises a probabilistic spatial wind speed distribution to determine turbine-level wind speeds. A rule-based methodology is applied to extract HWSS events from microscale wind speeds. The resulting binary event series is analysed using survival theory to create a timeto- event model for subsequent analysis and relative probabilistic comparison of risk between sites. AFRIKAANSE OPSOMMING: Die toenemende indringing penetrasie van windenergie het daartoe gelei dattot verandering in die gewone operasionele praktyke van konvensionele kragstelsels, veral stelsels wat deur termiese kragstasies oorheers word, hervorm word. Die veranderlike eienskappe aard van windenergie kragopwekking vereis dat verdere beleggings in meer buigsame aanlegte met vinniger aanlegvermoë aktiveringsvermoë gedoen gemaak moet word om teen moontlike tekorte aan kragin opwekking te beskerm. Hierdie stelsel vir vinnige reaksiedienste word deur die stelseloperateur ingeskakel na aanleiding van groot krag verhoogings toenamesveranderings om te verseker te maak dat die energiebalans gehandhaaf word. Vinnig reaksiedienste aanvaar neem tipies die vorm van berging of gasturbines, wat teen 'n hoë koste verkry word. In die konteks van die oprit van windkrag veranderingsverskynsels, is die hoë windsnelheidsuitsetting windsnelheidsafskakeling (HWSS) moontlik die ernstigste grootste risiko wat die stabiliteit van die kragstelsel kan beinvloed. Dit is duidelik vanuit die beskikbare literatuur dat HWSS tot op hede nog nie breedvoerig ondersoek is nie.Uit die beskikbare literatuur verskyn dit dat HWSS tot op hierdie punt nog nie volledig ondersoek is nie. Alhoewel die behoefte aan voorspelling en kwantifisering van die impak van HWSS sterk in die beskikbare literatuur voorkom, daar is daar tot dusver geen modelle geformuleer om hierdie verskynsel te beskryf nie, en daar is geen gebeurtenisgebaseerde -voorspellingsmodelle in antwoord op hierdie navorsingsvraag voorgestel nieontwikkeld. Hierdie proefskrif fokus op twee belangrike hoof aspekte van HWSS, naamlik die modellering, kwantifisering en vergelyking van die relatiewe risiko van HWSS-gebeure, en die korttermyn operasionele voorspelling van HWSS-gebeure. Uit die literatuur is dit duidelik dat die ontwikkeling van 'n pasgemaakte HWSS-voorspellingsmodel sal bydra bystand verleen totmet die bestuur en verligting vermidering van die korttermynrisiko wat met HWSS-gebeure verband hou. Die ontwikkeling van liggingspesifieke modelle waarmee tydelike tydgebonde risiko gekwantifiseer en vergelyk kan word, sal verder help met die plasing van windplase in gebiede met 'n lae waarskynlikheid vir HWSS-gebeure. Twee nuwe gebeurtenisgebaseerde gebeurtenisgebaseerde-voorspellingstegnieke word voorgestel vir die korttermynvoorspelling van HWSS-gebeure, naamlik 'n Kunsmatige Neurale Netwerk (ANN) - benadering, en 'n bastermodel hibriedmodel met behulp van 'n oorspronklike statistiese afskaleringafmetingsmetodologie. Beide van die voorgestelde modeltopologieë gebruik ‘n verskeidenheid windspoedvoorspellings wat afgelei is met behulp van die Weather Research and Forecasting (WRF) model, sowel as addisionele omgewingsveranderlikes soos windrigting en temperatuur. Die resultate bewys dat beide modelle 'n goeie akkuraatheid toon vir die voorspelling van 'n hoë windsnelheid gebeur afgelei op mikroskaalvlak, wat ooreenstem met die voorspelling van HWSS-gebeuregelokaliseerde gebeure met 'n hoë windsnelheid wat op mikroskaalvlak plaasvind. 'n Tegniek word voorgestel vir die area karakterisering en vergelyking van HWSS-gebeure vir die plasingsbeplanning van windplase. Die voorgestelde tegniek maak gebruik van 'n probabilistiese waarskynlike ruimtelike windsnelheidsverspreiding om windsnelhede op turbine-vlak te bepaal. 'n Reëlgebaseerde metodologie word toegepas om HWSS-gebeure uit mikroskaal-windsnelhede te onttrek. Die resulterende reeks van binêre opsies word met behulp van oorlewingsteorie geanaliseer om 'n tyd-tot-gebeurtenis-model te skep vir daaropvolgende ontleding en relatiewe probabilistiese waarskynlike vergelyking van risiko tussen windplaseareas. Doctoral 2019-11-20T13:08:09Z 2019-12-11T06:51:46Z 2019-11-20T13:08:09Z 2019-12-11T06:51:46Z 2019-12 Thesis http://hdl.handle.net/10019.1/107184 en_ZA 114 pages : illustrations application/pdf |
| spellingShingle | Winds -- Speed Wind farms Wind energy conversion systems Artificial neural network Groch, Matthew Modelling and short-term forecasting of high wind speed events at operational wind farms |
| title | Modelling and short-term forecasting of high wind speed events at operational wind farms |
| title_full | Modelling and short-term forecasting of high wind speed events at operational wind farms |
| title_fullStr | Modelling and short-term forecasting of high wind speed events at operational wind farms |
| title_full_unstemmed | Modelling and short-term forecasting of high wind speed events at operational wind farms |
| title_short | Modelling and short-term forecasting of high wind speed events at operational wind farms |
| title_sort | modelling and short term forecasting of high wind speed events at operational wind farms |
| topic | Winds -- Speed Wind farms Wind energy conversion systems Artificial neural network |
| url | http://hdl.handle.net/10019.1/107184 |
| work_keys_str_mv | AT grochmatthew modellingandshorttermforecastingofhighwindspeedeventsatoperationalwindfarms |