Impact of Measurement Uncertainties on Compressive Sensing-based Harmonic Source Estimation Algorithms

The harmonic pollution generated by non-linear loads and generators in distribution networks could lead to serious consequences for the customers. The identification of the polluting sources is thus necessary to act directly on the origin of the problems, in order to reduce the corresponding effects. As well as the other Power Quality phenomena, the harmonic analysis requires specific measurement devices, which are not widespread in distribution systems. Consequently, specific algorithms and methodologies are required, in order to overcome the lack of measurements. In this regard, Compressive Sensing (CS) represents a valid solution for the analysis under study, since it allows to recover sparse signals, when only few measurements are available.In this paper, two of the most common algorithms for CS applications, the Orthogonal Matching Pursuit and the ℓ
₁
-minimization, implemented in the framework of Harmonic Source Estimation procedures, are tested under different measurement uncertainty conditions. The tests, performed on a small medium voltage network, underline the different impact brought by the measurement uncertainties, of both magnitude and phase angle, on the two algorithms.