A Power Forecasting Method for Ultra-Short-Term Photovoltaic Power Generation Using Transformer Model

Table 3

Model parameters.

	Parameter	Transformer	GRU	DNN
Model parameter setting	Number of nodes	40	40	40
	Each batch sample size	32	32	32
	The number of iterations	100	100	100
	The steps of the sliding value	1	1	1
Model structure	Number of parameters of encoding layer (input layer)	571378	6600	560
	Number of hidden layer parameters	20	410	410
	Number of output layer parameters	11	11	11