Optimal Fault Predictors for Arc-Type Faults in Radial and Meshed Alternating Current Distribution Systems

Abstract

Efficient and reliable linear and nonlinear fault diagnostic systems are of vital importance in modern utility power grids as quick, accurate detection of faults can assist in preventing system failures that cause economic loss and endanger human or animal life. The foundation of any fault diagnostic system is its set of predictors; robust predictors naturally lead to a reliable system. This work fills the need for a deep investigation into reliable fault detection predictors. Novel harmonic-based fault predictors are developed for diagnosis of fault condition in both radial and meshed type AC distribution systems, with four fault classification groups (bolted fault, high impedance nonlinear fault, linear fault, and a no-fault classification). These new fault predictors are optimized and rigourously tested against earlier fault predictors using existing fault models and known statistical methods in both noiseless and noisy conditions.