Abstract
This paper presents a study on a possible topology of a novelty detector to be applied in Power Systems applications. In the context of Smart Grids, a novelty detector should not be limited to typical power quality (PQ) disturbances, but rather be capable of identifying any new deviation or variation in the signal. Therefore, traditional novelty detection methodologies based on a training stage are not suitable. The proposed detection approach divides the signal into fixed-length frames and relies on comparing the current frame with a reference frame using a specific metric. Results for two scenarios are presented, achieving a hit rate of up to 99.0% with a false alarm rate of 1.0%.