Robustness and Generalization of Machine Learning and Deep Learning Models for Power Quality Disturbance Classification

Authors

DOI:

https://doi.org/10.61799/2216-0388.2373

Keywords:

Power quality, power quality disturbances, machine learning, deep learning, cross-domain generalization, robustness, AWGN

Abstract

Automatic power quality disturbance classification is commonly evaluated under in-domain conditions, which may overestimate performance when data distributions or noise conditions change. This study evaluated the robustness and generalization capability of machine learning and deep learning models for the classification of eight power quality disturbance classes. Three data domains—SEED-PQD, PQD-16c, and a synthetic dataset—were harmonized to a common representation of one cycle, 64 samples, and a 3.2 kHz sampling rate. Support vector machines (SVM), Random Forest, XGBoost, 1D-CNN, and CNN-LSTM were compared using Macro-F1, balanced accuracy, MCC, cross-domain generalization, and degradation under additive white Gaussian noise from 30 to 0 dB. Deep learning models achieved in-domain Macro-F1 values above 0.98 on SEED and the synthetic domain when sufficient training data were available, although they showed high sensitivity to training-set size and noise. SVM achieved the highest average cross-domain Macro-F1 (0.6407) and the highest robustness AUC (0.4328) and retention AUC (0.5039). The Friedman test confirmed significant differences among models under noise (p<0.001; W=0.5265), whereas no statistically significant global differences were detected for cross-domain generalization (p=0.1257). The results show that high in-domain performance does not guarantee generalization or robustness and highlight the need to jointly evaluate classification performance, domain transfer, and noise tolerance before selecting models for real-world power quality applications.

Downloads

Download data is not yet available.

References

[1] A. A. Memon, M. A. Koondhar, S. F. Al-Gahtani, Z. M. S. Elbarbary, and Z. M. Alaas, “Comprehensive review of power quality disturbance detection and classification techniques,” Computers & Electrical Engineering, vol. 126, Art. no. 110512, 2025, doi: 10.1016/j.compeleceng.2025.110512.

[2] J. E. Caicedo, D. Agudelo-Martínez, E. Rivas-Trujillo, and J. Meyer, “A systematic review of real-time detection and classification of power quality disturbances,” Protection and Control of Modern Power Systems, vol. 8, Art. no. 3, 2023, doi: 10.1186/s41601-023-00277-y.

[3] J. C. Palomares-Salas, S. Aguado-González, and J. M. Sierra-Fernández, “Robustness of machine learning and deep learning models for power quality disturbance classification: A cross-platform analysis,” Applied Sciences, vol. 15, no. 19, Art. no. 10602, 2025, doi: 10.3390/app151910602.

[4] M. A. A. Baig, N. I. Ratyal, A. Amin, U. Jamil, S. Liaquat, H. M. Khalid, and M. F. Zia, “An ensemble deep CNN approach for power quality disturbance classification: A technological route towards smart cities using image-based transfer,” Future Internet, vol. 16, no. 12, Art. no. 436, 2024, doi: 10.3390/fi16120436.

[5] M. Mosayebi, S. Azad, and M. T. Ameli, “Unknown power quality disturbances classification based on transfer learning approach with imbalanced data considerations,” Results in Engineering, vol. 27, Art. no. 105865, 2025, doi: 10.1016/j.rineng.2025.105865.

[6] U. Sipai, R. Jadeja, N. Kothari, T. Trivedi, and K. K. Ram, “Deep transfer learning approach for the classification of single and multiple power quality disturbances,” Scientific Reports, vol. 15, Art. no. 34583, 2025, doi: 10.1038/s41598-025-18064-0.

[7] I. Kapuza, E. Ginzburg-Ganz, R. Machlev, and Y. Levron, “Improving robustness of Transformers for power quality disturbance classification via optimized relevance maps,” Engineering Applications of Artificial Intelligence, vol. 161, Art. no. 112138, 2025, doi: 10.1016/j.engappai.2025.112138.

[8] X. Zhang, C. Jiang, M. Yu, X. Wen, J. Zhang, N. Rong, and S. Han, “Adversarial black-box attack and defense for convolutional neural network-based power quality disturbance classification,” Engineering Applications of Artificial Intelligence, vol. 162, Art. no. 112411, 2025, doi: 10.1016/j.engappai.2025.112411.

[9] E. García Rodríguez, E. Reyes Archundia, J. A. Gutiérrez Gnecchi, O. I. Coronado Reyes, J. C. Olivares Rojas, and A. Méndez Patiño, “Detection and extraction of optimal features from power quality disturbances based on wavelet coefficient reconstruction and noise-robust methods,” Computers & Electrical Engineering, vol. 127, Art. no. 110615, 2025, doi: 10.1016/j.compeleceng.2025.110615.

[10] Q. Xu, F. Zhu, W. Jiang, X. Pan, P. Li, X. Zhou, and Y. Wang, “Efficient identification method for power quality disturbance: A hybrid data-driven strategy,” Processes, vol. 12, no. 7, Art. no. 1395, 2024, doi: 10.3390/pr12071395.

[11] M. U. Khan, S. Aziz, and A. Usman, “SEED_PQD_v1 (SEED—Power Quality Disturbance Dataset ver1),” Zenodo, dataset, 2024. [Online]. https://zenodo.org/records/11843312?utm

[12] M. U. Khan, S. Aziz, and A. Usman, “XPQRS: Expert power quality recognition system for sensitive load applications,” Measurement, vol. 216, Art. no. 112889, 2023, doi: 10.1016/j.measurement.2023.112889.

[13] jkexin, “PQD-16c-Data: Power quality disturbance dataset,” Kaggle, dataset. [Online]. https://www.kaggle.com/datasets/jkexin/pqd-16c-data?utm

[14] D. Li, I. A. Channa, X. Chen, L. Dou, S. Khokhar, and N. Ab Azar, “A new deep learning method for classification of power quality disturbances using DWT-MRA in utility smart grid,” Computers & Electrical Engineering, vol. 117, Art. no. 109290, 2024, doi: 10.1016/j.compeleceng.2024.109290.

[15] L. Chen, S. Chen, J. Xu, and C. Zhou, “Power quality disturbances identification based on deep neural network model of time-frequency feature fusion,” Electric Power Systems Research, vol. 231, Art. no. 110283, 2024, doi: 10.1016/j.epsr.2024.110283.

[16] T. Liao, W. Wang, and Y. Xing, “A method for disturbance identification in power quality based on cross-attention fusion of temporal and spatial features,” Electric Power Systems Research, vol. 234, Art. no. 110560, 2024, doi: 10.1016/j.epsr.2024.110560.

[17] P. Savaridass M and S. Esakkirajan, “Hilbert transform based combined 1D and 2D deep learning framework for power quality disturbance classification,” Electric Power Systems Research, vol. 249, Art. no. 112031, 2025, doi: 10.1016/j.epsr.2025.112031.

[18] M. A. A. Baig, N. I. Ratyal, A. Amin, U. Jamil, H. M. Khalid, and M. F. Zia, “A multi-modal deep learning framework for power quality disturbance classification: An integration of 1D time-series signals and 2D scalograms,” Computers & Electrical Engineering, vol. 128, Art. no. 110716, 2025, doi: 10.1016/j.compeleceng.2025.110716.

[19] P. Khetarpal, N. Nagpal, M. S. Al-Numay, P. Siano, Y. Arya, and N. Kassarwani, “Power quality disturbances detection and classification based on deep convolution auto-encoder networks,” IEEE Access, vol. 11, 2023, doi: 10.1109/ACCESS.2023.3274732.

[20] M. Liu, Y. Chen, Z. Zhang, and S. Deng, “Classification of power quality disturbance using segmented and modified S-transform and DCNN-MSVM hybrid model,” IEEE Access, vol. 11, 2023, doi: 10.1109/ACCESS.2022.3233767.

[21] M. Abubakar, A. A. Nagra, M. Mudassar, M. Faheem, and M. Sohail, “High-precision identification of power quality disturbances based on discrete orthogonal S-transforms and compressed neural network methods,” IEEE Access, vol. 11, 2023, doi: 10.1109/ACCESS.2023.3304375.

[22] V. Veeramsetty, A. Dhanush, A. Nagapradyullatha, G. Rama Krishna, and S. R. Salkuti, “Power quality disturbances classification using autoencoder and radial basis function neural network,” International Journal of Emerging Electric Power Systems, vol. 25, no. 6, pp. 817–842, 2024, doi: 10.1515/ijeeps-2023-0143.

[23] S. Janthong and P. Phukpattaranont, “Identification of power quality disturbances in electrical distribution system using fast Fourier transforms and super learner ensembles,” Journal of Advanced Research in Applied Mechanics, vol. 124, no. 1, pp. 39–60, 2024, doi: 10.37934/aram.124.1.3960.

[24] H. Zhang, W. Wu, K. Li, X. Zheng, X. Xu, X. Wei, and C. Zhao, “Multi-strategy active learning for power quality disturbance identification,” Applied Soft Computing, vol. 154, Art. no. 111326, 2024, doi: 10.1016/j.asoc.2024.111326.

[25] A. Zlatkova and D. Taskovski, “A novel CNN-based framework for detection and classification of power quality disturbances: Exploring multi-class versus multi-label classification,” IEEE Access, vol. 13, 2025, doi: 10.1109/ACCESS.2025.3546520.

[26] M. Alsabaan, A. Elsayed, A. Bondok, M. M. Badr, M. Mahmoud, T. Alshawi, and M. I. Ibrahem, “Robust federated-learning-based classifier for smart grid power quality disturbances,” Sensors, vol. 25, no. 22, Art. no. 6880, 2025, doi: 10.3390/s25226880.

[27] H. Castro-Gutiérrez, C. Robles-Algarín, and L. L. Camargo-Ariza, “Análisis comparativo de los algoritmos de árbol de decisión y máquinas de vectores de soporte en la clasificación de impulsores de bomba,” Respuestas, vol. 30, no. 1, pp. 95–106, 2025, doi: 10.22463/0122820X.5233.

[28] D. C. Candia-Herrera, K. Y. Sánchez-Mojica, S. L. Zarama-Ortiz, and C. L. Diaz-Mesa, “Hacia un modelo universal de IA para la detección simultánea de COVID-19, neumonía bacteriana y tuberculosis en imágenes de rayos X: Desafíos y oportunidades,” Respuestas, vol. 31, no. 1, pp. 47–61, 2026, doi: 10.22463/0122820X.5433.

[29] M. Á. Flórez-Zambrano, F. A. León-García, J. L. Díaz-Rodríguez, and O. J. Suárez-Sierra, “Sistema en tiempo real para el reconocimiento de vocales en LSC mediante CNN y visión por computador,” Respuestas, vol. 31, no. 2, 2026, doi: 10.22463/0122820X.5918.

Published

2026-09-01

Issue

Section

Artículo Originales

How to Cite

[1]
Cardozo Sarmiento, D.O. et al. 2026. Robustness and Generalization of Machine Learning and Deep Learning Models for Power Quality Disturbance Classification. Mundo FESC Journal. 16, 36 (Sep. 2026). DOI:https://doi.org/10.61799/2216-0388.2373.

Most read articles by the same author(s)