Mean-Shift Cross Energy Operator for Spectral Change Detection in Low Quality Harmonic Signals

The efficacy of power quality monitoring is influenced by data quality issues such as phase errors, bad data, etc. Specifically, harmonic change detection suffers with low-quality data and requires adaptive methods. Therefore, this letter proposes a novel mean-shift cross-energy operator for spectral change detection in the time domain with data-quality adaptability. The method quantifies spectral similarity between signal pairs as the mean difference of cross-energy without the influence of data errors. It is a powerful non-linear bivariate energy operator that extends the Teager cross-energy concept to multi-component signals, provides time-domain spectral change detection, and high dynamic sensitivity. For static snapshots of signals, the method is suitable in detecting change in harmonic composition or frequencies, irrespective of amplitude or phase, whereas during dynamic tracking, it captures changes in both the harmonic composition and amplitude/phase. It is tested with multitude of data quality issues in power measurements in Matlab/Simulink and with real-system data for ground current to distribution transformer at Kharagpur, India.