In the last years, the QuEEN monitoring system enhanced its functionalities integrating some innovative criteria to evaluate both the validity and origin of voltage dips by means of Deep Learning and Machine Learning techniques. In particular, the DELFI (Deep Learning for False voltage dip Identification) classifier obtained more promising results in the events validity assessment than those pursued by a first criterion based on the estimate of the 2nd harmonic.

In previous works of the authors, the influence of some networks characteristics on voltage dips performance have been evaluated by means of Multiple Linear Regression analysis (MLR) applied on events statistics whose validity were checked with the 2nd Harmonic Criterion. In this work, the MLR analyses have been performed on statistics validated with DELFI, i.e., with the data recorded by QuEEN in the periods 2018-2020 and 2021-2023.

In the MLR analysis the neutral operation, the geographical area and the overhead lines length were used as explicative variables, while three voltage dips performance indexes were adopted as response variables (i.e., total number of events in the period and total number of events occurred respectively under class 2 and class 3 immunity curve). Considering only the true events for DELFI, a greater influence of specific geographical area was observed on the counting indices, with a higher impact for the indices associated with greater severity.