Fault Detection and Classification in Power Grids Using Shannon Entropy with Convolution Distributed Autoencoder

Nowadays, Power grids have undergone significant transformations as the demand on electricity increased tremendously. Power grids are the primary sources for electricity generation and power distribution to the consumers need, but it faces numerous faults during this process. The traditional approaches for detection of faults in power grids had faced challenges, that includes accurately detecting and classifying which type of faults in the power grid. Therefore, this research proposes Shannon Entropy-Convolution Distributed Autoencoder (SE-CDAE) for fault detection and classification in power grids. The proposed method employed on Power Grid dataset which provides the current signals data that includes normal current signals and fault current signals. Then, the gathered data is pre-processed by fast wavelet transform which converts the time-domain signals into wavelet coefficient by removing the noise and dimensionality. These pre-processed data optimal features are extracted by using SE that effectively eliminates the irregular patterns in data. Finally, CDAE is employed for detecting and classifying the faults from extracted optimal features. The proposed SE-CDAE effectively detected and classified the faults in power grids and succeeded with high results including accuracy (97.98%), precision (94.83%), recall (96.28%), F1-score (97.35%) when compared with existing Semi-ProtoPNet Visual Geometry Group-19 (SPPN-VGG-19).