Using of neural network technology and multi-agent systems to preventing large-scale emergencies in electric power systems

Recent examples of large-scale blackouts in North American 2003, Moscow in 2005 and Europa in 2003 and 2006 have clearly demonstrated that secure operation of large interconnected power systems cannot be achieved without full understanding of the system behavior during abnormal and emergency conditions. This paper proposes an intelligent approach to the system monitoring and control with the goal of identification of potential voltage instability problems before they lead to major blackouts. The proposed approach is based on detecting alarm states using self-organized Kohonen neural networks, and activating a multi-agent control system to take necessary preventive actions. The Kohonen network is trained off-line and then applied on-line to predict possible emergencies. The proposed clustering model was realized in STATISTICA 8.0 and tested on the modified 42-bus IEEE power system. Results are presented and discussed.