Stacking machine learning model with enhanced DWT features for detection and classification of power quality disturbances in EV charging stations

The rapid increase in the use of electric vehicles has led to the establishment of charging stations across the power grid. Although electric vehicle charging stations provide significant social benefits and convenient access to charging services, their integration into the power grid may introduce challenges related to protection and power quality disturbances. Therefore, accurate detection of power quality disturbances is crucial to prevent equipment damage and ensure reliable operation. For this purpose, this paper proposed a framework based on an enhanced Discrete Wavelet Transform (DWT) for feature extraction and a stacking Machine Learning (ML) model to detect power quality disturbances in electric vehicle charging stations. In this paper, feature extraction has been performed using the DWT with the db4 smooth Mother Wavelet (MW) to mitigate the problem of high energy at the beginning and end of the time window in the DWT with the db4 MW, and enhance focus on the actual disturbance interval. Furthermore, a stacking ML model has been proposed to improve overall detection accuracy. To assess robustness, the proposed framework is further evaluated under multiple practical variations, including different noise levels, sampling rates, DWT decomposition levels, and real-world data from a charging station The proposed framework achieved an outstanding detection and classification accuracy, particularly in the presence of complex power quality disturbances. The results are compared to various state-of-the-art methods and evaluated using multiple evaluation metrics, confirming the reliability and excellent performance of the proposed approach.