Open AccessAutomated Tool Based on Deep Learning to Assess Voltage Dips Validity: Integration in the QuEEN MV network Monitoring System
Author(s) -
Michele Zai,
R. Chiumeo,
Liliana Tenti,
Massimo Volta
Publication year - 2021
Publication title -
renewable energy and power quality journal
Language(s) - English
Resource type - Journals
ISSN - 2172-038X
DOI - 10.24084/repqj19.265
Subject(s) - voltage , computer science , queen (butterfly) , classifier (uml) , power quality , artificial intelligence , data mining , real time computing , engineering , electrical engineering , hymenoptera , botany , biology
This paper presents the development of an automated tool called QuEEN PyService, aimed to the extraction of events voltage signals from the QuEEN distribution network monitoring system database, for advanced Power Quality analysis. The application has allowed the integration of the DELFI classifier (DEep Learning for False voltage dips Identification), recently developed by RSE, making it possible for the first time the intensive validation of the latter on a large number of voltage dips. Thanks to this tool, a comparison between the performance of DELFI and those of an older criterion based on the 2nd voltage harmonic measurement has been performed using data recorded by 61 measurement units in the period 2015-2020 The analysis has been focused on traditional PQ voltage dips counting indices as N2a e N3b. Results show that the usage of the DELFI classifier increases the N2a and the N3b by respectively the 20.6 % and 38.8% with respect to the QuEEN criterion.