Table 9.

Performance comparison of different multi-label classification NILM methods in seen scenarios. The number in bold is the largest of the three model comparisons.

Device	Model	F1score	Precision	Recall	Accuracy	MCC	MAE	SAE
AHU0	CNN	0.871	0.783	0.982	0.907	0.812	273.71	0.286
	TP-NILM	0.926	0.875	0.985	0.948	0.891	169.85	0.160
	TTRNet	0.892	0.986	0.819	0.938	0.859	185.19	−0.148
AHU1	CNN	0.683	0.529	0.971	0.800	0.612	702.30	0.836
	TP-NILM	0.772	0.650	0.959	0.872	0.716	484.95	0.485
	TTRNet	0.951	0.978	0.928	0.979	0.940	142.76	−0.059
AHU2	CNN	0.871	0.781	0.985	0.894	0.798	422.14	0.206
	TP-NILM	0.923	0.873	0.984	0.938	0.879	272.84	0.088
	TTRNet	0.992	0.998	0.986	0.994	0.988	80.41	−0.057
AHU5	CNN	0.880	0.795	0.985	0.902	0.813	1235.75	0.325
	TP-NILM	0.930	0.874	0.996	0.945	0.891	847.91	0.223
	TTRNet	0.991	0.990	0.992	0.994	0.986	380.09	0.070