Clustering and Fusion of Distributed Multitone Frequency Measurement Data for Enhanced Grid-Level Power Quality Monitoring

The increasing diffusion of both nonlinear loads and renewable-based distributed energy resources (DERs) may cause serious power quality (PQ) problems due to the injection of significant harmonics and interharmonics. To mitigate some of these problems, the controllers of smart power electronic converters (PECs) can embed techniques for harmonics’ compensation. However, to maximize their performance, the most relevant narrowband disturbances should be detected and their frequency estimated with high accuracy. Given that harmonics and interharmonics maintain consistent frequencies across different points in the grid, this article introduces a flexible technique that leverages multiple distributed estimates to accurately identify critical narrowband disturbances propagating throughout the system. The proposed approach relies on four main steps: the initial two are designed for implementation in various measurement devices to identify local interferers, while the latter two are intended to operate on a central aggregator. This central unit collects the multitone distributed frequency and amplitude estimates to cluster and fuse data corresponding to the same interferers. A key novelty of the proposed approach lies in the clustering algorithm that relies on an adaptive threshold depending on the actual measurement uncertainty of the available data. This method effectively filters out potential outliers, reducing the risk of misdetections, before calculating the uncertainty-weighted average of frequency values within each cluster. Several Monte Carlo (MC) and grid-level simulations confirm the excellent performance of the proposed approach in identifying critical sinusoidal disturbances, thus paving the way for better distributed compensation strategies to be implemented in PECs.