Open Access

ARTICLE

DC Disturbance Classification Method Based on Compressed Sensing and Encoder

Huanan Yu¹, Xiang Zhang^1,*, Jian Wang²

1 Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Ministry of Education (Northeast Electric Power University), Jilin, 132012, China
2 Iron law Energy Company, Liaobei Technician College, Liaoning, 112700, China

* Corresponding Author: Xiang Zhang. Email:

(This article belongs to the Special Issue: Advanced Analytics on Energy Systems)

Energy Engineering 2025, 122(12), 5055-5071. https://doi.org/10.32604/ee.2025.067152

Received 26 April 2025; Accepted 18 August 2025; Issue published 27 November 2025

Abstract

Recent advances in AC/DC hybrid power distribution systems have enhanced convenience in daily life. However, DC distribution introduces significant power quality challenges. To address the identification and classification of DC power quality disturbances, this paper proposes a novel methodology integrating Compressed Sensing (CS) with an enhanced Stacked Denoising Autoencoder (SDAE). The proposed approach first employs MATLAB/SIMULINK to model the DC distribution network and generate DC power quality disturbance signals. The measured original signals are then reconstructed using the compressive sensing-based generalized orthogonal matching pursuit (GOMP) algorithm to obtain sparse vectors as the final dataset. Subsequently, a Stacked Denoising Autoencoder model is constructed. The Root Mean Square Propagation (RMSprop) optimization algorithm is introduced to fine-tune network parameters, thereby reducing the probability of convergence to local optima. Finally, simulation analyses are conducted on five common types of DC power quality disturbance signals. Both raw signals and sparse vectors are utilized as datasets and fed into the encoder model. The results indicate that this method effectively reduces the feature dimensionality for DC power quality disturbance classification while improving both recognition efficiency and accuracy, with additional advantages in noise resistance.

---

Keywords

---