A Power Transformer Fault Prediction Method through Temporal Convolutional Network on Dissolved Gas Chromatography Data

Table 6

Gas regression results of transformer no. 1.

Gas	Metrics	MSTCN	TCN	LSTM	GRU
C2H2	RMSE	0.0051	0.0026	0.0065	0.0042
	MAE	0.0043	0.0020	0.0055	0.0036
	MAPE	14.07%	5.80%	17.96%	11.66%
		0.7091	0.9254	0.5218	0.7975
C2H4	RMSE	0.0056	0.0058	0.0089	0.0065
	MAE	0.0045	0.0045	0.0071	0.0052
	MAPE	1.32%	1.31%	2.05%	1.51%
		0.8503	0.8393	0.6161	0.7967
C2H6	RMSE	0.0494	0.0567	0.0768	0.0721
	MAE	0.0450	0.0491	0.0646	0.0630
	MAPE	1.24%	1.33%	1.76%	1.73%
		0.9419	0.9237	0.8599	0.8764
CH4	RMSE	0.0138	0.0200	0.0284	0.0350
	MAE	0.0116	0.0155	0.0230	0.0314
	MAPE	0.41%	0.54%	0.81%	1.10%
		0.9294	0.8516	0.7019	0.5483
CO	RMSE	12.2925	12.3431	13.9061	16.6197
	MAE	8.8837	10.1510	11.9101	13.2390
	MAPE	2.03%	2.26%	2.65%	2.87%
		0.8095	0.8079	0.7562	0.6518
CO2	RMSE	44.8718	50.1873	47.7348	53.6491
	MAE	31.9983	40.4018	40.1977	41.3562
	MAPE	2.56%	3.11%	3.07%	3.25%
		0.8856	0.8569	0.8705	0.8364
H2	RMSE	0.0083	0.0108	0.0535	0.0468
	MAE	0.0062	0.0095	0.0524	0.0458
	MAPE	0.13%	0.20%	1.12%	0.98%
		0.8810	0.7981	−3.9776	−2.8074