Self-tuning Kalman filter and machine learning algorithms for voltage dips upstream or downstream origin detection

In this paper, self-tuning Kalman Filter (KF) is applied to a significant sample of full waveforms associated to the voltage dips monitored in the Italian distribution network by the QuEEN system, with the aim of events detection and waveforms segmentation. Segmentation is done in order to extract more features and information from the original voltage waveforms, to make easier voltage dips classification, based on the events source location (upstream/downstream from the point of measurement). The aforementioned classification is achieved by Machine Learning algorithms. The evaluation of the obtained results is based on the computation of a “confusion matrix”.