AI Driven Approach to Power Quality Detection and Classification

Power quality events present significant challenges to the stability and reliability of electrical grids, necessitating efficient detection methods for timely response. This enables the rapid implementation of diagnostic and mitigation methods to address disturbances quickly. Traditional approaches to detecting power quality events depend on predefined thresholds or models, which may lack the flexibility needed for diverse and evolving grid conditions. In contrast, data-driven approaches use machine learning techniques to automatically identify patterns from extensive real-world data, offering greater flexibility and accuracy. This paper introduces power quality event detection methods based on machine learning techniques. Utilizing models such as eXtreme Gradient Boosting (XGBoost), Support Vector Machines (SVM), Random Forest (RF), Neural Networks (NN), and Convolutional Neural Networks (CNN), the approach effectively identifies and classifies various power quality events, including voltage sags, swells, and interruptions. The study showcases the superiority of these AI methods in enhancing the detection process and facilitating timely mitigation strategies. A Simulink simulation model is used to evaluate the performance of various machine learning algorithms, with XGBoost and CNN demonstrating superior performance over other models. Overall, the findings underscore the potential of AI methods in improving the detection and prediction of power quality events, contributing to more resilient and adaptable electrical grids.