Unified Quantile Regression Deep Neural Network with Time-Cognition for Probabilistic Residential Load Forecasting

Table 1

Shared parameters of neural networks and related strategies selected.

Parameters	LSTM	CNN
Depth	3	34
Hidden neural	128	None
Kernel size	None	8
Kernel number	None	24
Batch size	128	128
Input size	240	240
Dropout rate	0.05	0.05
Loss function	Quantile loss	Quantile loss
Optimizing method	AMSGrad	AMSGrad
Start learning rate	0.01	0.01
Learning rate decay	0.3	0.3
Training stop	Early stopping	Early stopping