Fig 1.
The suggested system block diagram.
Fig 2.
Flow chart of the MASA.
Fig 3.
The features extraction principle.
Fig 4.
The instances for pure signal, sag, swell and intr.
Fig 5.
Zooms of instances of sag and intr.
Fig 6.
The instances acquired with a 4–bit resolution SPADC for sag (left) and intr. (right).
Fig 7.
Number of samples per instance for sag (left) and swell (right).
Table 1.
Summary of the samples ratios.
Table 2.
Accuracy scores for power signals recognition, obtained by different classifiers.
Table 3.
Performance of considered classifiers using F1, AUC and Kappa metrics on the test dataset.
Table 4.
Confusion matrices for the KNN and the NB classifiers.
Table 5.
Confusion matrices for the SVM and the ANN classifiers.
Table 6.
Comparison with state-of-the-art methods.