Power Dispatch Automation System for Fault Prediction and Diagnosis Based on Long Short-Term Memory and Spatial-AlexNet

Now a days, the rapid evolution of electrical networks has given rise to the smart grid, a next-generation network that has captured the attention of the scientific community. However, smart grids are largely limited to basic threshold-based approaches using single-system data, indicating a need for more sophisticated solutions. This research proposes a method for faults prediction and diagnosis in automation power distribution. Initially, the input data is collected from the dataset of power forecasting then, preprocessing is performed using the Short-Time Fourier Transform (STFT) which is a time-frequency energy mapping method remove the noise from input data. Then the features are selected by Time series Generative Adversarial Networks (Time GAN) whereas it captures patterns and for the generation of real time series. Finally, the prediction is done through Long Short-Term Memory with Spatial-AlexNet (LSTM-S-AlexNet), it detects the faults and handles complex patterns to select the best structure for accurate diagnosis. From the results, the proposed LSTM-S-AlexN et shows better performance in terms of accuracy of (98.56%), precision of (99.38%), recall of (97.84%) and F1-Score of (98.42%) when compared to existing method of Bidirectional Sliced Gated Recurrent Unit with Gated Attention mechanism (BiSGRU-GA) respectively.