A hybrid model for wind speed prediction in Anuradhapura, Sri Lanka

Abstract

Wind energy plays a major role in a sustainable future as a useful, environmentally friendly energy alternative. Wind speed is the most important parameter in the design and implementation of wind energy. This paper aims to define a methodology capable of providing accurate monthly average wind speed predictions in the Anuradhapura region, Sri Lanka. Hybrid forecasting of time series is considered to be a potentially effective alternative compared with the conventional stand-alone forecasting modeling approaches like seasonal autoregressive integrated moving average (SARIMA) and artificial neural network (ANN). In this study, at first, SARIMA and ANN models are used to separately recognize and forecast the linear and nonlinear components of time series, respectively. Then, the study suggests a hybrid approach combining SARIMA and ANN for forecasting wind speed and its forecasting results are compared with the single SARIMA and ANN models. The mean absolute error (MAE), root mean square error (RMSE), and paired sample t-test are used as performance measures. Results obtained by a case study show that the SARIMA-ANN hybrid approach is the most suitable for wind speed forecasting. This approach demonstrates the potential to be applied to wind speed forecasting in other regions of the country.