Forecasting Wind Speed in Abu Dhabi Using Statistical Modelling and Machine Learning Approaches

Abstract

Wind energy is a sustainable energy source which is considered a reliable alternative source of energy that develops very fast all over the world. Forecasting wind speed is a global concern and a critical issue for wind power conversion systems as it has a great influence in scheduling the power systems as well as the dynamic control of wind turbines, thus, it is desired to have an accurate wind speed forecasting. There are various number of research papers that studied wind speed prediction and forecasting using so many different methods or models. These methods are divided into four categories; physical methods, statistical methods, spatial correlation methods, and machine learning methods. In this research thesis, we will deploy and study four forecasting models in order to forecast wind speed data of Abu Dhabi city. The models are divided into two categories, two from statistical methods (ARIMA and SARIMA models) and the other two from machine learning methods (ANN and SSA models). Afterwards, we will compare these four models in order to determine the most efficient model in forecasting the wind speed data. The results of our study showed that the forecasting model SSA had the most accurate forecasted values comparing to the other three models, however, it also showed that those three models, ARIMA, SARIMA and ANN, had batter results for the first few hours (around 24 hours), which indicates that ARIMA, SARIMA, and ANN models are suitable for short-term forecasting, while SSA model are suitable for long-term forecasting. The findings of our research could contribute in defining the fitting forecasting model in terms of short-term forecasting or long-term forecasting.

Date of Award	Dec 2016
Original language	American English
Supervisor	U Zeyar Aung (Supervisor)