Comparison and Optimization of Neural Networks and Network Ensembles for Gap Filling of Wind Energy Data

Table 13

Summary of results from the 3 best neural networks out of 20000 iteratively tested networks for the target turbines in the complex terrain of the Sauerland region.

Target variable	Input variables	Network type	Network topology	RMSE (kWh)	𝑟𝑆
𝑊WE1	𝑊(OL1--OL3)	MLP	3-9-1	640	0.90
𝑊WE1	𝑊(OL1--OL3)	RBF	3-22-1	602	0.91
𝑊WE1	𝑊(OL1--OL3)	RBF	3-22-1	588	0.92
𝑊WE2	𝑊(OL1--OL3)	MLP	3-7-1	747	0.92
𝑊WE2	𝑊(OL1--OL3)	RBF	3-24-1	630	0.93
𝑊WE2	𝑊(OL1--OL3)	RBF	3-21-1	668	0.90