Novelty Detection in Power Quality Signals with Surrogates: a Time-Frequency Technique

Abstract

In the context of smart grids (SGs), new types of electrical disturbances may arise in current and voltage waveforms of electrical power systems (EPSs). Detecting and analyzing such disturbances is important for the operation and control of EPSs. To identify waveform abnormalities or variations, novelty detectors can be employed. This work presents a novelty detection technique for power quality signals based on surrogate signals and distance measures computed in the time-frequency domain. Surrogates are generated by preserving the magnitude of the Fourier spectrum of the original signal while randomizing the phase spectrum using a uniform distribution. Tests were conducted using synthetic signals representing common power quality disturbances. The proposed method is compared with an energy-based novelty detector available in the literature, showing improved performance in terms of receiver operating characteristic (ROC) curve metrics. Additionally, a real-world waveform signal is used to demonstrate the effectiveness of the proposed technique. The surrogate-based approach can also be useful for defining ground truth in novelty detection tasks involving real power system signals.