Fig. 1: Comparison of the next 24 h of WSP by each algorithm.
From: A machine learning model for hub-height short-term wind speed prediction
a is RMSE result. b is MAE result. c is ACC result. For a fair comparison with the ECMWF numerical model, the machine learning 24-hprediction model is used here to predict twice a day, each time taking the results of the first 12 h. MFWPN−6 denotes that the MFWPN 24-h prediction model predicts four times a day, each taking the results of the first six h of the prediction. The breaks on the horizontal axis indicate the intervals where the predictions are made twice a day. These figures shows that MFWPN outperforms ECMWF in the first 6 h of wind prediction.
