Outage predictions of electric power systems under Hurricane winds by Bayesian networks

Abstract

The stability and reliability of electric power grids are essential to the continuous operation of modern cities as well as for the mitigation, preparedness, response and recovery in disaster management. Power systems must be assessed in order to identify and address component and system-level weaknesses while supporting their rapid restoration. This paper proposes a Bayesian Network (BN)-based framework to predict outages in an electric power grid that is exposed to a hurricane event. The inherent capabilities of BNs, including its intuitive and graphical representation of probabilistic information, and its ability to allow information updating with ease, make it an effective tool for this application. The framework is coupled with a DC-flow model that captures the physics of the electrical system and also reduces the computational complexity of building conditional probability tables needed in the BN model. The framework relies on component fragilities and topology of the grid, and predicts outages at substations and distribution points within the electric power system. Its application is demonstrated using Harris County's electric power system under the 2008 Hurricane Ike winds that battered the Gulf Coast of the United States. The model captured well the field system response, where low outage probabilities are observed in the transmission system while the outage risks at distribution load points are significantly higher. The developed BN framework can seamlessly integrate transmission and distribution systems, inform disaster management, and suggest restoration strategies, while supporting decision-making for pre- or post-event intervention actions.