Classification of Power Quality Disturbances (PQDs) Analyzed Using Deep Learning Techniques

Abstract

Power Quality (PQ) issues in distributed generation primarily arise from excessive nonlinear loads inside the system. Identification and categorization are essential to guarantee the reliability of Power Quality Disturbances (PQDs). This paper presented a signal processing and deep learning methodology to categorize Power Quality Disturbances (PQDs) utilizing Discrete Wavelet Transform (DWT), Multi-Resolution Analysis (MRA), and a one-dimensional Convolutional Neural Network (CNN). To expedite training, the performance of the model utilized signal processing-based DWT-MRA to extract 54 features, which were subsequently input into a 1D-CNN. The implementation of 1D-CNN appears to be more dependable than alternative machine learning methodologies. Simulation results demonstrated effective performance and efficient data classification. Consequently, the suggested methodology may herald a new epoch for PQDs in photovoltaic/wind smart grids, yielding more efficient results in the near future.