Research Article
An Improved Power Quality Disturbance Detection Using Deep Learning Approach
Table 10
Comparison with the existing methods.
| Methods | Number of features selected | Number of PQDs | Accuracy in % | Pure | 40 dB | 30 dB | 20 dB |
| WPT and GA [18] | 15 | 8 | 98.33 | - | - | - | WT and PSO [19] | 11 | 9 | 98 | 96.87 | 93.625 | – | ST and RF [30] | 4 | 15 | 99.7 | 99.9 | 99.7 | 95.9 | FFT and ANNs [31] | – | 8 | – | 93.95 | 95.65 | – | ST and PNN [32] | 4 | 11 | 97.4 | - | - | - | FDST and DT [33] | 20 | 13 | 99.28 | 98.8 | 97.49 | – | Deep CNN [24] | AUTO | 9 | 99.67 | - | - | - | Deep CNN [25] | AUTO | 16 | 99.96 | 99.95 | 99.66 | 98.13 | Deep CNN [26] | AUTO | 24 | - | - | - | 99.26 | The proposed CNN + LSTM | AUTO | 16 | 100 | 100 | 99.69 | 99.31 |
|
|