Document Type : Original Article

Authors

• Mojtaba Shivaie ¹
• Mahdi Rahnamaei Hashjin ¹
• Felipe Vigolvino Lopes ²

¹ Shahrood University of Technology

² Federal University of Paraíba (UFPB)

Abstract

Power quality disturbances (PQDs), including voltage sags, swells, harmonics, and transient fluctuations, pose significant challenges to the stability, reliability, and efficient operation of large-scale industrial distribution grids, particularly with the increasing integration of photovoltaic (PV) systems. This paper presents a deep learning-based approach for the detection and classification of PQDs using real and simulated data from an active industrial distribution network. The distribution grid is modeled and simulated in ETAP 19.0.1, incorporating power load-flow and harmonic analyses to represent the operating conditions of the studied system. A deep neural network (DNN) is trained using a comprehensive dataset to classify the predefined PQD categories. The proposed DNN achieves a training accuracy of approximately 94% and a validation accuracy of approximately 92%, compared with the 85% accuracy reported for the DAG-SVM approach developed by Li et al. [3]. The results demonstrate that the proposed approach can effectively classify PQDs and provide a scalable framework for real-time power quality monitoring. Furthermore, the proposed method enhances the capability of industrial distribution systems to identify and manage disturbances under diverse operating conditions. These findings highlight the potential of deep learning as a practical and scalable solution for PQD classification and monitoring in modern industrial power systems, particularly in PV-integrated environments.

Keywords

• Active industrial distribution network
• deep learning
• power quality disturbances
• real-time power quality monitoring

Main Subjects

• Application of AI and machine learning