Robust Fault Detection of Wind Energy Conversion Systems Based on Dynamic Neural Networks

Table 2

The training and testing results for the network structure selection to model 𝜔_𝑔.

Network structure	𝐾	Training			Testing
		𝐽	𝑓FPE	𝑓AIC	𝐽	𝑓FPE	𝑓AIC
N^22,3,1(1)	25	0.0282	0.0296	−3.5148	0.0126	0.0129	−4.3491
N^22,4,1(1)	33	0.0379	0.0405	−3.2068	0.0200	0.0207	−3.8790
N^22,5,1(1)	41	0.0258	0.0286	−3.5554	0.0151	0.0157	−4.1521
N^22,6,1(1)	49	0.0521	0.0575	−2.8566	0.0573	0.0602	−2.8105
N^22,7,1(1)	57	0.0406	0.0455	−3.0900	0.0664	0.0703	−2.6551
N^22,3,1(1)	31	0.0268	0.0285	−3.5574	0.0134	0.0138	−4.2815
N^22,4,1(1)	41	0.0417	0.0453	−3.0953	0.0424	0.0442	−3.1196
N^22,5,1(1)	51	0.0378	0.0419	−3.1734	0.0409	0.0430	−3.1456
N^22,6,1(1)	61	0.0349	0.0394	−3.2333	0.0327	0.0348	−3.3594
N^22,7,1(1)	71	0.0795	0.0917	−2.3900	0.0850	0.0913	−2.3941
N^32,3,2,1(2-2)	50	0.1286	0.1421	−1.9510	0.0365	0.0384	−3.2604
N^32,4,2,1(2-2)	61	0.0902	0.1019	−2.2837	0.0767	0.0815	−2.5069
N^32,4,3,1(2-2)	73	0.1135	0.1314	−2.0300	0.0141	0.0152	−4.1886