Short-term wind speed prediction using various forecasting methods

Nambandi, Maria Ndinelago (2020-03)

Thesis (MCom)--Stellenbosch University, 2020.

Thesis

ENGLISH SUMMARY : There is a significant challenge in finding ways to enhance energy security and decrease greenhouse gas emissions emanating from the consumption of non-renewable resources for energy. The release of greenhouse gases causes global warming and is considered not clean. Compared to current conventional sources of energy, such as fossil resources, renewable energy sources have become more attractive for electricity production as it has been identified as clean with a closed carbon dioxide cycle. Thus, the CO2 produced during processing is reabsorbed by plants for food production. A significant development in the electricity industry in recent years has been the fast growth of wind power. The wind power generated from the wind depends on meteorological conditions such as wind speed and wind direction. These meteorological conditions are considered stochastic in nature, especially wind speed, and attempts to accurately forecast future values are therefore considered important in power generation. There are various studies in the literature which make use of statistical techniques to predict wind speed data. In this thesis, the short-term prediction of hourly wind speeds at 60m hub height for two sites, Jozini and Memel in South Africa, over a 1 to 24-hour forecast horizon is considered. The potential short-term wind speed at a site was predicted using statistical forecasting techniques such as traditional time series models (ARMA, ARIMA, seasonal ARIMA and regression using Fourier terms with ARMA errors), multilayer perceptron (MLP) neural networks and long short term memory (LSTM) recurrent neural networks. These predictions are relevant for planning purposes to ensure that the necessary base load on the electricity grid is established at all times. All forecasting techniques were applied to forecast wind speeds for each forecast horizon and site. Different LSTM and MLP configurations were created using a different number of hidden layers, a different number of hidden nodes in each layer, different learning rates, and different activation functions. The forecast performances of each configuration were compared to the persistence forecast (benchmark model). Root mean square errors (RMSE) and mean absolute percentage errors (MAPE) were used to select the configuration that best predicted the test data. Our empirical results show that the three different statistical techniques considered achieved similar results for each site and all the forecast horizons. Seasonal ARIMA models were used because there was a clear indication that the wind speeds data for Jozini and Memel are seasonal, with daily and annual regular cycles respectively. For the Memel site, a more accurate model was obtained through the use of regression with ARMA errors, where the Fourier term corresponding to annual seasonality was used as a regressor. Overall, the persistence forecast was the least accurate model to predict wind speeds at 60m height for all sites. LSTM configurations and regression using Fourier terms with ARMA errors achieved similar results with a slight improvement for the latter. Neural networks achieved comparable results with traditional time series models, thus, suggesting that the behaviour of wind speeds for each site is not overly complicated and simple forecasting techniques can be used for modelling. The predicted values obtained using the most accurate model, and the actual values for each site were plotted. The results showed that regression with Fourier terms and ARMA errors method could accurately predict the oscillations of the wind speed series with high accuracy, and it predicted most of the sudden peaks in the series. The analysis reported in this work provides much insight into wind speed forecasting for researchers who might apply statistical forecasting techniques on wind data in the future.

AFRIKAANSE OPSOMMING : Daar is ‘n groot uitdaging om maniere te vind om energiesekuriteit te bevorder en kweekhuisgasse, wat veroorsaak word deur die verbruik van nie-hernubare energiebronne, te verminder. Die vrylating van kweekhuisgasse veroorsaak aardverwarming en word nie as skoon beskou nie. In vergelyking met die konvensionele bronne van energie, soos fossielbronne, word hernubare energie as meer aantreklik beskou aangesien dit as skoon geidentifiseer is en omdat dit oor ‘n geslote koolstofdioksied siklus beskik. Dus word die CO2 wat deur hierdie proses gegenereer word deur plante geabsorbeer vir voedselproduksie. ‘n Belangrike ontwikkeling in die elektrisiteitsindustrie die afgelope paar jare is die vinnige groei van windenergie. Die energie wat uit wind gegenereer word is afhanklik van meteorologiese toestande soos windspoed en windrigting. Hierdie meteorologiese toestande, veral windspoed, is stogasties in natuur en die vooruitskatting van waarnemings word dus as belangrik beskou in kragopwekking. Daar is verskeie studies in the literatuur wat gebruik maak van statistiese tegnekie om windsnelhede te voorspel. In hierdie tesis word korttermyn vooruitskatting van uurlikse windspoedmetings by ‘n 60m naaf hoogte vir twee liggings, Jozini en Memel in Suid-Afrika, oor ‘n 1 tot 24 uur vooruitskattingshorison oorweeg. Die korttermyn windspoed by ‘n ligging was voorspel deur gebruik te maak van statistiese vooruitskattingstegnieke soos tradisionele tydreeks modelle (ARMA, ARIMA, seisoenale ARIMA en regressie met Fourier terme en ARMA foute), veelvoudige vlak “perceptron” (MLP) neurale netwerke en lang- korttermyn geheue (LSTM) herhalende neurale netwerke. Hierdie voorspellings is relevant vir beplanningsdoeleindes om te verseker dat die nodige basislading op die elektrisiteitsnetwerk verseker word. Alle vooruitskattings tegnieke was gebruik om windsnelhede te voorspel vir elke kombinasie van vooruitskattingshorison en ligging. Verskeie LSTM en MLP samestellings was geskep deur verskillende hoeveelhede verskuilde vlakke, hoeveelhede verskuilde nodusse, leerkoerse en aktiveringsfunksies te oorweeg. Die vooruitskattingskwaliteit van elke samestelling was vergelyk met die volhardingsvooruitskatting (maatstaf model). Die wortel gemiddelde kwadraat fout (RMSE) en die gemiddelde absolute persentasie fout (MAPE) was gebruik om die samestellings te kies wat die toetsdata die beste voorspel. Ons empiriese resultate toon aan dat die drie verskillende statistiese tegnieke wat oorweeg was soortgelyke resultate lewer vir elke kombinasie van ligging en vooruitskattingshorison. Seisonale ARIMA modelle was gebruik omdat daar ‘n duidelike aanduiding is dat die windsnelhede vir Jozini en Memel seisonaal is, met ‘n daaglikse en jaarlikse siklus onderskeidelik. 'n Meer akkurate model was verkry vir die Memel ligging deur 'n regressie met ARMA foute model te pas waar die Fourier term met 'n jaarlikse periode ooreenstem. In die algemeen was die volhardingsvooruitskatting die minste akkuraat om windsnelhede by by ‘n 60m hoogte te voorspel. LSTM samestellings en regressie met Fourier term en ARMA foute het soortgelyke resultate gelewer, met die laasgenoemde wat ‘n effense verbetering getoon het. Neurale netwerke en tradisionele tydreeksmodelle het soortgelyke resultate gelewer wat aandui dat die gedrag van windsnelhede nie oormatig kompleks is nie en gemodelleer kan word deur eenvoudige vooruiskattingstegnieke. Vir die toetsdata was die voorspelde waarnemings van die mees akkurate model saam met die werklike waarnemings gestip. Die resultate toon aan dat regressie met Fourier terme en ARMA foute die ossillasies in die windsnelhede akkuraat kan voorspel. Hierdie model was ook in staat om meeste van die skielike pieke in die tydreekse te herken. Die analise wat in hierdie werkstuk uitgevoer is bied insig in windspoedvooruitskatting vir navorsers wat statistiese tegniek op winddata in die toekoms wil toepas.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/107757
This item appears in the following collections: