Contents lists available at ScienceDirect Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr Three-phase voltage events classification algorithm based on an adaptive threshold Jorge L. Strack a,b,c, ⁎ , Ignacio Carugati a,b,c , Carlos M. Orallo a,b,c , Sebastián O. Maestri a,b,c , Patricio G. Donato a,b,c , Marcos A. Funes a,b,c a Instituto de Investigaciones Científicas y Tecnológicas en Electrónica (ICYTE), Argentina b Universidad Nacional de Mar del Plata (UNMdP), Juan B. Justo 4302, Mar del Plata, Argentina c Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Av. Rivadavia 1917, Buenos Aires, Argentina ARTICLE INFO Keywords: Voltage dip Voltage swell Event classification Power quality meter ABSTRACT A new algorithm for voltage event classification based on predefined models is proposed in this paper. In par- ticular, this proposal is designed for the ABC classification criterion and, unlike other algorithms based in the same criterion, this is based on adaptive classification thresholds that allows to reduce the classification errors in the presence of changes in the grid parameters (impedance, prefault voltage, etc.). It is demonstrated that the proposal presents a better performance than other algorithms described in the specific bibliography when are tested with real voltage events that differ from ideal three-phase model. A comprehensive study of algorithm classification errors considering several disturbances and experimental results are also presented. 1. Introduction Electrical disturbances such as voltage dips, swells and interruptions are of particular interest because they are very common in electrical grids [1]. They are usually classified as voltage events, all of which have different causes including line faults [2], induction motors start [3], transformers energization [4,5], etc. In order to evaluate, quantify, characterize and mitigate them, it is essential to have appropriate tools to detect and classify events in an automated way, with the shortest time and the greatest possible accuracy. When refer to classification the idea is to identify and group voltage events with similar characteristics using a set of predefined event models. That can be as complex or simple as necessary. It would be useful to have models that represent the types of faults that are most commonly found or that are most re- presentative. There are different ways of approaching this issue, but the more cited in the specific bibliography are the symmetrical components and ABC criteria [1]. The second criterion will be adopted in this work, but many of the concepts developed here can be easily extrapolated from ABC criterion to symmetrical components criterion. Three algo- rithms for voltage event classification are outlined in the specific bib- liography, known as SCA (Symmetrical Components Algorithm) [6], SPA (Six-Phases Algorithm) [6] and SVA (Space Vector Algorithm) [7,8]. They allow to properly classify ideal voltage events according to the ABC classification criterion [1]. Nevertheless, the authors demonstrated in [9] that they fail under the presence of different dis- turbances. Such as phase jumps and phase rotations, due to the im- pedance characteristics of the grid, as well as load and fault im- pedances, and by the use of arbitrary classification thresholds when the pre-fault voltage is different from the nominal value. In order to reduce these classification errors, in this paper a new algorithm called Absolute Sequences Algorithm (ASA) is proposed. This algorithm uses the in- formation provided by the absolute value of the three fundamental symmetrical components of the three-phase voltage to evaluate which is the more probable event type that matches the measured voltages. The proposal does not use fixed classification thresholds, which allows to reduce the classification error when the pre-fault voltage differs from the nominal value. It adapts to the amplitude and phase variations of the voltages during the fault. The improvement of the proposed algo- rithm with regard to the performance of the other algorithms described in the bibliography is evidenced with different tests. In addition, the capacity for event type discrimination of the proposed algorithm and its sensitivity to noise present in the power grid are evaluated. In Section 2 is described the classification criterion used, and in Section 3 are characterized the main disturbances that may take place in conjunction with a voltage event, which may produce events mis- classification. In Section 4 is described the proposed algorithm and its performance is analyzed in Section 5. Its performance is evaluated by a comparison with other methods under the same conditions. In Section 6 https://doi.org/10.1016/j.epsr.2019.03.012 Received 26 November 2018; Received in revised form 25 February 2019; Accepted 14 March 2019 ⁎ Corresponding author at: Universidad Nacional de Mar del Plata (UNMdP), Juan B. Justo 4302, Mar del Plata, Argentina. E-mail address: jlstrack@fi.mdp.edu.ar (J.L. Strack). Electric Power Systems Research 172 (2019) 167–176 0378-7796/ © 2019 Elsevier B.V. All rights reserved. T