Wind Speed Forecasting with a Clustering-Based Deep Learning Model

The predictability of wind energy is crucial due to the uncertain and intermittent features energy. This study proposes speed forecasting models, which employ time series clustering approaches deep learning methods. (LSTM) model utilizes preprocessed data as input returns features. Dirichlet mixture dynamic time-warping method cluster time-series then in forecasting. Particularly, warping Next, models use entire (global) clustered (local) capture long-term short-term patterns, respectively. Furthermore, an ensemble obtained by integrating global local results exploit advantages both models. Our are tested on four different from locations Turkey with regimes geographical aspects. numerical indicate that proposed achieve best accuracy compared (LSTM). imply feature approach accommodates a promising framework