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

Table 7

Classification metrics of different models.

Transformers	Metrics	MSTCN	TCN	LSTM	GRU
Transformer no. 1	Precision	99.20%	98.59%	98.55%	98.55%
	Recall	99.12%	98.29%	98.24%	98.24%
	F1-score	0.9914	0.9837	0.9831	0.9831
Transformer no. 2	Precision	99.33%	99.08%	98.87%	98.95%
	Recall	99.18%	98.76%	98.35%	98.53%
	F1-score	0.9922	0.9885	0.9850	0.9865
Transformer no. 3	Precision	98.67%	98.40%	98.23%	98.21%
	Recall	98.06%	97.41%	96.94%	96.88%
	F1-score	0.9823	0.9770	0.9733	0.9728
Transformer no. 4	Precision	98.16%	98.03%	97.94%	97.84%
	Recall	97.12%	96.76%	96.53%	96.24%
	F1-score	0.9741	0.9713	0.9694	0.9671