Methodology for a Security/Dependability Adaptive Protection Scheme Based on Data Mining

Abstract

Recent blackouts offer testimonies of the crucial role played by protection relays in a reliable power system. It is argued that embracing the paradigm shift of adaptive protection is a fundamental step toward a reliable power grid. The purpose of this paper is to present a methodology to implement a security/dependability adaptive protection scheme. The advocated methodology aims to reduce the likelihood of manifestation of hidden failures and potential cascading events by adjusting the security/dependability balance of protection systems. The proposed methodology is based on wide-area measurements obtained with the aid of phasor measurement units. A data-mining algorithm, known as decision trees, is used to classify the power system state and to predict the optimal security/dependability bias of a critical protection scheme. The methodology is tested on a detailed 4000-bus system.