A Simple Method of Residential Electricity Load Forecasting by Improved Bayesian Neural Networks

Table 7

Results of time-series BNN and BNN_16 model.

Model			Training set (60%)			Test set (20%)			Computing time (s)
	Time delay	Hidden neurons	MSE	R	MAPE	MSE	R	MAPE
					(%)			(%)
TS_BNN	1	2	1.07e-1	8.58e-1	11.75	1.05e-1	8.57e-1	11.86	6
	2	2	1.06e-2	8.58e-1	11.79	1.04e-1	8.63e-1	12.25	5
	50	2	8.79e-2	8.83e-1	10.74	8.76e-2	8.87e-1	11.34	192
	1	6	1.01e-2	8.65e-1	11.46	9.80e-2	8.74e-1	11.67	27
	1	8	9.84e-2	8.68e-1	11.73	1.05e-1	8.66e-1	11.50	29
BNN_16	-	2	9.27e-2	8.78e-1	11.38	9.41e-2	8.75e-1	11.40	2
	-	3	8.90e-2	8.84e-1	11.41	9.65e-2	8.68e-1	11.55	2
	-	5	8.56e-2	8.88e-1	10.93	9.46e-2	8.75e-1	10.72	3
	-	8	8.26e-2	8.92e-1	10.64	8.56e-2	8.87e-1	10.52	6
	-	10	8.16e-2	8.93e-1	10.54	8.77e-2	8.88e-1	10.92	7
	-	15	7.78e-2	8.99e-1	10.55	8.97e-2	8.80e-1	10.42	18
	-	18	7.72e-2	8.99e-1	10.36	8.95e-2	8.83e-1	10.87	139
	-	20	7.62e-2	9.01e-1	10.43	8.80e-2	8.83e-1	10.94	148
	-	30	7.19e-2	9.07e-1	10.13	9.29e-2	8.78e-1	10.98	355
	-	50	6.47e-2	9.17e-1	9.62	9.41e-2	8.73e-1	11.59	732
	-	100	5.81e-2	9.26e-1	9.04	1.07e-1	8.59e-1	12.50	1856