Integration of Higher-Order Time-Frequency Statistics and Neural Networks: Application for Power Quality Surveillance

Abstract

Higher-order statistics demonstrate their innovative features to characterize power quality events, beyond the traditional and limited Gaussian perspective, integrating time-frequency features and within the frame of a Higher-Order Neural Network (HONN). With the massive advent of smart measurement equipment in the electrical grid (Smart Grid), and in the frame of high penetration scenarios of renewable energy resources, the necessity dynamic power quality monitoring is gaining even more importance in order to identify the suspicious sources of the perturbation, which are nonlinear and unpredictable in nature. This eventually would satisfy the demand of intelligent instruments, capable not only of detecting the type of perturbation, but also the source of its origin in a scenario of distributed energy resources.