•  
  •  
 

Keywords

power quality disturbance signal; two-dimensional convolutional neural network; spectrogram; depth-separable convolution; disturbance signal recognition; lightweight neural network

Abstract

In response to difficulties in ac curately classifying and recognizing complex power quality disturbance (PQD) signals, this paper proposes a novel PQD recognition method based on spectrogram and lightweight two-dimensional (2D) depth-separable convolutional neural network (2D-DSCNN). Time-frequency analysis is applied to convert PQD signals into spectrograms, so that complex signal data is presented in the form of images. A lightweight 2D-DSCNN model is constructed by using depth-separable convolution technology, and the spectrograms corresponding to different PQD signals are classified and identified. The feasibility and effectiveness of the proposed method are verified through simulation experiments. The experimental results show that the method can effectively recognize and classify various PQD signals with high accuracy and strong anti-noise capability, and the model is lightweight, which is suitable for the deployment of edge devices and real-time monitoring.

DOI

10.19781/j.issn.1673-9140.2025.06.014

First Page

147

Last Page

155

Download

Share

COinS