Combining Digital Signal Processing, Artificial Intelligence and Graphs for Modeling Power Quality Disturbances

Abstract

This article presents the combination of digital signal processing, artificial intelligence and acyclic graphs for modeling power quality disturbances. These techniques include continuous wavelet transform, fuzzy logic and Bayesian networks. The aim with these techniques is to detect and characterize power quality disturbances. Later, by using Bayesian networks it is possible to diagnose the source of the disturbance within a previously modeled power system. These techniques are implemented into an instrument for measuring power quality disturbances. The performance of this instrument is evaluated in laboratory by injecting it power signals with known disturbances generated for representing several scenarios. The results reveal that the developed instrument can detect a wide spectrum of power quality disturbances. Additionally, this system can diagnose the cause of the disturbance to support corrective actions and infrastructure improvements over the analyzed power system.