Probabilistic Short-Term Wind Power Forecasting Using Sparse Bayesian Learning and NWP

Table 6

Power forecast error of WF-J (67 wind turbines, 100.5 MW).

Model	RMSE (MW)	MAE (MW)	NMAPE	NMAPE improvement
Persistence	26.34	18.22	18.13%	—
MLP-CS	17.32	12.04	11.98%	33.92%
SVM-CS	17.27	12.00	11.94%	34.14%
SBL-CS	16.46	11.42	11.36%	37.34%
OSBL-CS	15.29	10.58	10.53%	41.92%