Correction to “Novel Probabilistic Neural Network Models Combined with Dissolved Gas Analysis for Fault Diagnosis of Oil-Immersed Power Transformers”

When we were reading our published article, we found an inconspicuous little data error. The reason for the error is that when we d matlab to simulate the IEC three-ratio method, the numerical input in the code was wrong, resulting in inaccurate experimental data of the IEC three-ratio method. The IEC three-ratio method was the only traditional method introduced in this paper, which was used to compare and highlight the novel artificial intelligence fault diagnosis method proposed in this paper, so this error does not affect the conclusion of the full text. We rewrote the matlab code for the IEC three-ratio method and obtained new experimental results for the IEC three-ratio method under the exact same experimental conditions (including the same computer, the same version of matlab, and the same fault diagnosis data set, etc.). The new experimental data of the IEC three-ratio method have no influence on the previous conclusions. We apologize for our oversight.

The following are the updated experimental data for the IEC three-ratio method (abbreviated as IEC in the table) in Table 7, Table 8, and Table 9.

Table 7. Comparison Results of Different Methods.

	accuracy (%)
fault type	IEC
LT (<150 °C)	0.30
LT (150–300 °C)	100.00
PD	100.00
AD	100.00
ave	75.08

Table 8. Efficiency and Error Rate of Different Methods.

method	time (s)	error rate (%)
IEC	2.1702	63.24

Table 9. Mean Square Error of Different Methods.

algorithms	MSE of train sample	MSE of test sample
IEC	0.6511	0.4727