Learning Processes to Predict the Hourly Global, Direct, and Diffuse Solar Irradiance from Daily Global Radiation with Artificial Neural Networks

Table 3

Best performance of learning processes for different FFNN configurations.

Model	Input matrix	Hidden neurons	R	RMSE (kWh/m^2)
1	SD, HA, GD, Kt	15	0.980	0.0707
2	SD, GD, Kt	10	0.846	0.1224
3	Sf, HA, GD, Kt	15	0.953	0.0707
4	Sf, HA, Kt	12	0.892	0.1140
5	SD, HA, Kt	10	0.920	0.0836
6	SD, HA, GD	10	0.891	0.1090
7	δ, SD, HA, GD, Kt	20	0.984	0.0390
8	δ, SD, HA, GD, Kt, Wss	10	0.950	0.0707