The Forecasting of Wind Speeds Derived from the Horizontal Scanning Doppler Lidar and a Tall Tower Situated at Ragged Point, Barbados Cloud Observatory Using Various Machine Learning Techniques

The 2-minute temporal resolution wind speeds from a horizontal scanning Doppler LIDAR situated at Ragged Point, Barbados Cloud Observatory, will be average over an interval of 10 minutes. Windspeed forecasts will be made for time forecast horizons of h = 6, 12, 18, 24, 30, 36 or 1, 2, 3, 4, 5, 6 hours ahead. The forecasting model framework that will be utilized in this study is not restricted to Transformer Networks. These results will be compared to linear statistical modeling such as Autoregressive Integrated Moving Average (ARIMA) models as well as the benchmark method of persistence. The models forecasts will be plotted along with the actual series and the metrics inclusive of RMSE and MAE will be evaluated for each h and all models under consideration. This project and its results are useful for short term forecasting applicable to wind farm economic load dispatch planning and load reasonable decisions amongst others.