Development of automatic intelligent system for on-line voltage security control of power systems

The majority of recent large-scale blackouts have been caused by voltage instability. A prompt on-line assessment of voltage stability for preventive corrective control of electric power systems is one of the key objectives for Control centers. The use of classical approximation methods alone is complicated. Therefore, several modified methods combined with machine learning algorithms enabling security assessment in real time have been proposed over the last years. The paper presents an automatic intelligent system for on-line voltage security control, which is based on the model of decision trees Proximity Driven Streaming Random Forest (PDSRF). In this case, the combination of original properties of PDSRF and capabilities of L-index as a target vector makes it possible to provide the functions of dispatcher warning, localization of critical nodes, and ensure direct interaction with the security automation systems. The efficiency of the proposed system was demonstrated using various test schemes of IEEE