Metodologia de apoio à operação de sistemas elétricos de potência baseada em identificação de sistemas com aprendizado de máquina

Abstract

Recent advances on synchro phasor measurement along with its regulation by the Operador Nacional do Sistema – ONS brought access to high resolution data with rates up to one sample per cycle. Electrical power system current, voltage, phase and frequency are collected through Phasor Measurement Units-PMUs and stored in Phasor Data Concentrators – PDCs. Such high-resolution data favours new algorithms and applications towards electrical power systems dynamic behaviour, once unattainable due to supervisory control and data acquisition system limitations. Implementations regarding real time operations help control room operators’ awareness and reactions under electrical power system disturbances and malfunctions. System identification methods, together with artificial intelligence tools such machine learning, contribute to research and development on real time supervisory control. In this study a classic system identification algorithm is implemented to process dynamic electrical power system data and find a model representing its behaviour through PMU data. A stabilization diagram is applied as a post processing method and results used to build a histogram underlining electrical power systems electromechanical modes.