Developing a Local Neurofuzzy Model for Short-Term Wind Power Forecasting

Table 5

RMSE improvement* achieved by the proposed LLNF method with respect to the compared approaches.

Test month	Persistence method	Multivariate ARIMA	RBF	MLP	RNN	LSSVMs	LLNF
April 2010	73.5%	61.1%	40.3%	36.2%	29.2%	33.1%	32.4%
May 2010	62.6%	57.3%	38.6%	48.7%	27.1%	29.5%	23.4%
June 2010	61.9%	57.4%	49.6%	42.6%	31.6%	23.3%	30.1%
July 2010	72.2%	66.7%	54.1%	50.7%	38.8%	37.8%	49.5%
Average	67.3%	60.6%	46.8%	44.7%	31.5%	33.7%	34.2%

∗RMSE improvement = (RMSE of compared method − RMSE of LNF)/(RMSE of compared method) × 1.