Short-Term Solar Irradiance Prediction Based on Multichannel LSTM Neural Networks Using Edge-Based IoT System

Table 1

Single feature data prediction result evaluation index.

	10 min					30 min					60 min
	MAE	RMSE		SMAPE	MAPE	MAE	RMSE		SMAPE	MAPE	MAE	RMSE		SMAPE	MAPE
Bagging	30.1	68.24	0.942	13.74	19.32	106.2	132.9	0.9	22.4	38.28	46.28	90.2	0.896	23.8	75.1
MLP	30.9	66.7	0.943	15.82	44.02	44.15	97.23	0.91	27.5	59.46	47.47	88.4	0.902	24.4	62.2
LSTM	33	68.25	0.943	18.61	24.89	60.4	97.14	0.895	27.6	57.13	127.4	97.1	0.895	27.6	57.13
BiLSTM	34.1	68.25	0.944	19.06	28.67	51.54	91.1	0.901	24.7	64.12	51.54	91.1	0.901	24.7	64.12
CNN-LSTM	32.6	67.65	0.944	15.15	22.72	51.14	81.11	0.861	30	53.38	51.14	81.1	0.861	30	53.38
CNN-BiLSTM	30.1	66.62	0.946	15.14	18.72	48.04	82.74	0.912	26	49.69	48.04	82.7	0.912	26	49.69
WT-LSTM	22.9	31.24	0.987	8.78	11.83	30.79	40.71	0.982	13.7	25.13	24.57	35.7	0.985	15.7	26.11
WT-BiLSTM	16.6	23.17	0.994	9.33	11.39	19.46	27.66	0.991	12.8	25.47	25.05	34.1	0.985	12.1	24.16
Proposed	10.1	13	0.998	8.94	9.3	11.38	17.86	0.996	10.8	15.24	29.48	38.4	0.98	13.1	21.82