Wind Speed Event Forecasting using a Hybrid WRF and ANN Model

This work forms a prelude to the development of an effective high wind speed shutdown short-term forecasting tool. A cold-start weather research and forecasting model is coupled with an Artificial Neural Network (ANN) to forecast wind events at a wind farm. The forecaster is aimed at the estimation of the number of turbines in a wind farm exceeding two wind speeds thresholds, i.e. 12 ms
^-1
and 18 ms
^-1
, for a 24 hour forecasting horizon. Two metrics are targeted in the forecast, namely the peak magnitude of an event, and the number of turbines exceeding the threshold wind speed in an event. It is shown that the model performs well for the 12 ms
^-1
threshold, and improves for an increasing percentage of turbines exceeding the threshold wind speed. For the 18 ms
^-1
threshold, an intermediary ANN model is introduced to improve the forecast in the wind speed range of interest. The improved model results in a good high wind speed event forecaster. It is evident from the results that the accuracy of the numerical forecasting model is of paramount importance at higher wind speeds.