Early warning system for spatiotemporal prediction of fault events in a power transmission system

Abstract

Predictions of the spatiotemporal distribution of fault events help the management planning and the maintenance scheduling in power transmission systems. To this end, this study proposes an early warning system based on the exploration of cause-attribute relationships. A comprehensive assessment of all the fault causes is conducted by incorporating entire available environmental attributes in the investigated system as inputs. To cope with the rarely occurred fault causes and environmental elements, a procedure of the weighted association rule mining is established. Then, a risk evaluation model is employed in evaluating the prediction performance and the risks of possible false predictions. Three requirements for the uncertainty are taken into account when this framework is applied in real use: Technical reliability, internal reliability, and factitious effectiveness. An empirical study is conducted in a real power transmission system to verify the feasibility of the proposed framework, and the results prove that a flexible and robust prediction can be generated consequently.