This document presents the integration of advanced machine learning techniques into the QuEEN monitoring system of the Italian distribution network. This system aims to monitor voltage dips in the Italian distribution network primarily for assessment and research purposes. For each recorded event, the system can automatically evaluate the residual voltage and duration based on the corresponding effective voltage values, and assign each event a “validity” (invalidating false events caused by voltage transformer saturation) and an “origin” (upstream/downstream from the measurement point) using appropriate procedures and algorithms (current techniques). In recent years, however, RSE has proposed new solutions to improve the algorithms for determining the validity and origin of events: the DELFI classifier (DEep Learning for False voltage dips Identification) and the FExWaveS + SVM classifier (Features Extraction Waveform Segmentation + Support Vector Machine). These advanced functionalities have been recently integrated into the monitoring system through the QuEEN PyService software tool.
In this work, advanced techniques were applied intensively for the first time across a significant number of monitored sites (150), using data recorded between 2018 and 2021. Moreover, a comparison between innovative techniques and current criteria was conducted at the macro-regional level. The new techniques have shown a significant impact on the number of severe voltage dips, confirming a non-uniform distribution of these events across different Italian macro-regions.