Wind speed forecasting using First Order Markov Transition with Regime Switch and Time Duration

Due to the dynamic nature of time series, it is common to deploy varying models in predicting behavior of the inherent characteristics in time series. The growing need in nonlinear series prediction and the limitations of linear models in predicting the behaviors of such, lead to the quest of models capable of predicting both dynamic and non-linear behaviors in time series. Markov Regime switching model commonly referred to as the regime switching model is used to forecast wind speed taking into account variation of speed with time. Eviews, a statistical, econometric and economic modelling package is used for the Markov regime switch estimation, simulation and forecasting of A 24-hour period real wind speed data measurement from NREL. Values obtained from forecast accuracy evaluation are Root Mean Squared Error (RMSE) is at 0.242, the Mean Absolute Error (MAE) is 0.17 and the Mean Absolute Percentage Error (MAPE) is 31.38. The statistics showed that the forecast is closely related to the actual values.