Power Quality Monitoring System Using Spectrogram 1D-CNN

This report presents a novel approach for identifying harmonic and non-harmonic disturbances in power quality monitoring systems using spectrogram analysis and a one-dimensional convolutional neural network (1DCNN). The proposed method uses advanced signal processing techniques to extract temporal and spectral features from power signals, enabling an accurate classification of disturbances such as sags, swells, harmonics, and oscillatory transients. The 1DCNN architecture is optimized to recognize complex patterns in power signals. The integration of spectrogram-based prior processing and deep learning not only enhances the reliability of power quality monitoring but also provides detectable faults for real-time applications in modern power systems.