GridGuard: Intelligent Load Monitoring and Fault Diagnosis in Renewable-Powered Smart Grids

Smart grids (SG) are important components of modern energy infrastructure, integrating renewable sources with intelligent monitoring to ensure reliability and efficiency. SG integrates renewable energy sources with real-time monitoring system and adaptive control to deliver reliable power generation and optimized grid efficiency. In this study, we propose a nine-layer deep learning architecture for real-time load forecasting and fault classification using a specific dataset. The dataset contains 50 k inputs at 15-minute intervals and captures key electrical parameters. This study addresses fault overload condition classification and load prediction (in kW). The study undergoes data preprocessing; imbalanced training data are handled using the synthetic minority oversampling technique (SMOTE). The proposed model handles regression and classification tasks simultaneously and complex temporal constraints effectively. The classification model effectively distinguishes fault states with high accuracy, while the regression model achieves excellent predictive performance, boasting an R2 score of 0.9942, an MAE of 0.1746, and an MSE of 0.0517 on the test set. The key findings demonstrate the potential of proposed model to study complex interactions, providing significant improvements in fault detection accuracy. The combination of computation and experiments in an integrated framework, this study demonstrates excellent analytical capabilities in smart energy systems, supporting accurate decisions and dynamic energy management.