Support-Vector-Machine-Based Proactive Cascade Prediction in Smart Grid Using Probabilistic Framework

The worldwide major blackout events of power network are highlighting the need for technology upgradation in traditional grid. One of the major upgradations required is in the area of early warning generation in case of any grid disturbances such as line contingency leading to cascade failure. This paper proposes a proactive blackout prediction model for a smart grid early warning system. The proposed model evaluates system performance probabilistically, in steady state and under dynamical (line contingency) state, and prepares a historical database for normal and cascade failure states. A support vector machine (SVM) has been trained with this historical database and is used to predict blackout events in advance. The key contribution of this paper is to capture the essence of the cascading failure using probabilistic framework and integration of SVM machine learning tool to build a prediction rule, which would be able to predict the scenarios of the blackout as early as possible. The proposed model is validated using the IEEE 30-bus test-bed system. Proactive prediction of cascade failure using the proposed model may help in realizing the grid resilience feature of smart grid.