A Power Quality Event Classifier Based on Convolutional Neural Network

This article introduces a novel and accurate automatic classifier for Power Signal (PS) alterations. The efficient detection and accurate classification represent crucial steps in developing an automatic power quality (PQ) measurement system, which has become essential in contemporary settings. The objective of this article is to enhance alteration classification accuracy. This is accomplished by combining the Hilbert-Huang transform (HHT) and convolutional neural network (CNN). The HHT is employed to extract PS features and is resilient to the non-stationarity introduced by alterations. Meanwhile, the CNN is adept at extracting information from the bidimensional characteristics of PS features and is robust against noise. Numerical tests demonstrate promising results showing high classification accuracy also in the case of PSs affected by high level of noise.