Spectral Continuity and Subspace Change Detection for Recovery of Missing Harmonic Features in Power Quality

The disturbance monitoring, event sequence recording, and automated fault analysis in the power system require processing of power quality and digital fault recorder data. Through, reliable and granular data streaming/storage services, which inadvertently introduces unwanted data quality issues like data gaps or missing samples. The work proposes a data-driven, gap length, and spectral change independent missing harmonic recovery method using static and dynamic selectivity criteria of spectral continuity and rate of subspace affinity change. For static selectivity, a novel mean-shift cross-energy operator is proposed that quantifies the spectral similarity between the static snapshots of signals across the gap. For dynamic selectivity, a novel rate of subspace change method is proposed to detect the subspace change points in a dynamically changing data set. Based on the selectivity criterion, the missing harmonic parameters are estimated and filled using the rotational invariance technique. The proposed method could effectively reconstruct the power signals with longer data gaps, contiguous, and randomly gaped data sets under dynamic harmonic conditions. The proposition is tested with simulated data sets in Matlab/Simulink and real system data from the India grid.