Threshold selection in diagnosis approach applied to power systems

Fault detection has been becoming a significant topic especially in the control of large scale complex systems. The goal is to detect faults as early as possible and provide a timely warning. Faults are detected using thresholds on residuals generated from differences between sensors measurements and computed values. The selection of thresholds affects the performance of the detection system. In this work we propose a method concerning the thresholds selection that can be used for alarm activation using the methodology of System Operating Characteristic (SOC) curves. The method is tested on an electric power transmission system presented in our previous work.