A novel decomposition-based localized short-term tidal current speed and direction prediction model

To integrate tidal energy into the existing grid, it is important to accurately predict the amount of tidal energy available in the short-term horizon. In this regard, the short-term prediction of tidal current becomes crucial The tidal energy depends not only on the tidal current speed (TCS), but also on the tidal current direction (TCD). The non-stationarity and non-linearity of the TCS and TCD time series lower the predictability of them. Using decomposition approaches, these non-linear and non-stationary time series can be decomposed into several components which are more predictable. On the other hand, in order to predict any volatility of a non-linear time series, localized approaches perform better compared to training the prediction model in a global fashion. In this regard, this paper proposes a novel prediction model based on ensemble empirical mode decomposition (EEMD) and localized least squares support vector machine (LSSVM) to increase the prediction accuracy of TCS and TCD. The proposed model is compared with the auto-regressive integrated moving average (ARIMA) model and the LSSVM-based model. The actual data recorded from the Shark River Entrance (Florida, U.S.) is used to verify the applicability of the proposed prediction model.