Clustering‐based method for the feeder selection to improve the characteristics of load shedding. Issue 4 (30th September 2019)

• Record Type: Journal Article
• Authors: Potel, Barnabé; Cadoux, Florent; De Alvaro Garcia, Leticia; Debusschere, Vincent
• Is Part Of: IET smart grid. Volume 2:Issue 4(2019)
• Page Start: 659
• Page End: 668
• Publication Date: 2019-09-30
• Subjects: pattern clustering -- load shedding -- distribution networks; medium‐voltage feeders -- given frequency threshold -- given feeder -- given step -- UFLS target -- allocation method -- France -- automated clustering technique -- introduced performance metrics -- described methods -- distribution‐side photovoltaic generation -- chosen performance metrics -- clustering‐based method -- feeder selection -- load shedding -- under‐frequency load -- frequency thresholds -- electrical power system -- generation‐load imbalance -- UFLS step; B0260 Optimisation techniques -- B8110B Power system management, operation and economics -- B8120J Distribution networks; Smart power grids -- Periodicals; Computer science -- Periodicals; Energy industries -- Periodicals; Broadcasting -- Periodicals; 333.79110285
• DOI: 10.1049/iet-stg.2019.0064
• Languages: English
• ISSNs: 2515-2947

Abstract

Abstract : Under‐frequency load shedding (UFLS) schemes are designed by specifying a given amount of load to shed at various frequency thresholds to prevent the collapse of the electrical power system in the event of a large generation‐load imbalance. An UFLS step is constituted of a group of medium‐voltage feeders that trip when a given frequency threshold is reached. This study focuses on the method to be used when allocating a given feeder to a given step. First, the authors introduce performance metrics to quantify the accuracy level with which the UFLS target is met. Second, they model: the allocation method currently used in France; a variant of that method; and a new method introduced in this study, based on an automated clustering technique. Third, based on real consumption patterns measured from a vast area in France, and using the introduced performance metrics, they compare the efficiency of the three described methods. This study is conducted for the current state of loading of the considered distribution network and for a hypothetical situation with an increased share of distribution‐side photovoltaic generation. For the chosen performance metrics, they demonstrate that the first two methods provide similar results while the clustering‐based method performs remarkably better.