By: Ravikumar Chinthaginjala, Asadi Srinivasulu, Anupam Agrawal, Tae Hoon Kim, Sivarama Prasad Tera & Shafiq Ahmad
Originally published in a science reportJuly 15, 2025.
Abstract
This study presents a new approach to improving power quality using semiconductor devices by integrating machine learning (ML), deep learning (DL), and advanced control strategies. This study addresses major power quality challenges, including voltage sagging, swelling, harmonics, and temporary turbulence, through a data-driven framework that combines traditional control techniques with adaptive learning models. Various algorithms including support vector machines (SVM), random forests, neural networks, convolutional neural networks (CNNs), and long-term memory (LSTM) networks were tested using real-time data. The results showed significant differences in performance in deep learning models, particularly LSTM, which proved to be more accurate and reliable when identifying and predicting power quality problems. In contrast, traditional ML models such as SVM and Random Forest have difficult class imbalances, resulting in reduced accuracy and recall. However, the DL model managed these challenges effectively, with CNN achieving 91.8% accuracy, and LSTM achieving full accuracy (100%) and 94.5% recall. This study highlighted the complications of handling disproportionate datasets, as indicated by the classification warning, and highlighted the importance of improved preprocessing and model adjustment for reliable predictions. Runtimes vary widely, and traditional control systems are faster but less capable of identifying complex patterns compared to computationally intensive DL models. These findings highlight the promise of hybrid systems that integrate both traditional and data-driven control strategies to achieve adaptive and reliable power quality control. Both simulations and real-world experiments support the effectiveness of this hybrid method, suggesting a strong foundation for intelligent power quality solutions in future smart grid applications. In this study, deep learning models provide excellent accuracy and predictive power for complex power quality scenarios, but practical deployment requires careful balance of computational demands and address class distribution challenges.
introduction
Power quality is a fundamental aspect of the stability and reliability of modern power systems1. Sub-power quality can cause a variety of problems, including sagging voltage, swelling, harmony distortion, and temporary disturbances, which can negatively affect industrial machinery, household electronics and sensitive equipment.2,3,4,5. Traditional power quality control methods often rely on static control strategies and preset thresholds.6. With the integration of renewable energy sources and increasing demand for electricity, there is an increasing need for advanced technologies that can adapt to these rapid changes.7. This study explores how machine learning (ML), deep learning (DL), and control algorithms can be integrated with semiconductor devices to create sophisticated power quality management systems. By combining the adaptability of data-driven models with the reliability of traditional control systems, this study proposes a hybrid approach to improve monitoring, fault detection, prediction, and aggressive control of power quality failures.8. The purpose of this study is to overcome the limitations of conventional techniques and highlight the advantages of adaptive algorithms for managing complex, real-time power quality data.9.
To continue reading this article, click here.
