Artificial intelligence and machine learning techniques for power quality event classification: a focused review and future insights

Abstract

Power Quality (PQ) disturbances are critical in modern power systems, significantly impacting electrical networks' stability, reliability, and efficiency. With the increasing penetration of renewable energy sources, non-linear loads, and power electronic devices, the detection, classification, and mitigation of PQ disturbances have become more complex. Traditional PQ analysis methods, which rely heavily on human expertise and rule-based systems, are often insufficient in handling the growing complexity and volume of data in real-time applications. This review comprehensively analyzes the latest advancements in Artificial Intelligence (AI) and Machine Learning (ML) techniques applied to PQ analysis, achieving classification accuracies as high as 99.94 % with hybrid approaches like dual-tree wavelet packet transforms combined with extreme learning machine (ELM). Integrating advanced signal processing techniques, such as wavelet transforms and empirical mode decomposition, has demonstrated accuracy improvements of up to 5 % in challenging scenarios. This paper explores the challenges associated with AI-based PQ analysis, including the need for large datasets, overfitting issues, and the lack of interpretability in complex models. Future research directions are outlined, emphasizing the development of hybrid models, explainable AI systems, and real-time adaptability to dynamic grid conditions. This review provides a holistic understanding of state-of-the-art AI/ML methods in PQ analysis. It highlights their potential to transform modern power systems by ensuring higher reliability, better fault detection, and more efficient power delivery.