Microcontroller-Based Residential Microgrid Power Management with IoT Integration

Resumo

Microgrids combine battery energy storage devices and renewable resources with the utility grid. An intelligent energy switching approach between solar, battery and grid ensures continuous supply of power by dynamically adjusting to resource availability and load requirements. In this paper, a novel microcontroller-based energy management scheme (MCU-based EMS) is proposed by combining both on-grid and off-grid power management systems allowing microgrids to function smoothly in both grid-connected and islanded modes. In the off-grid management, a Maximum Power Point Tracking (MPPT) algorithm is used for maximizing the extraction of available PV power. Further, a DC-DC boost converter is employed for boosting the voltage to match the battery’s charging need and loads. In on-grid management, an H-inverter along with a synchronizer circuit is used to supply loads during peak shaving hours. Besides, a bidirectional buck-boost converter ensures battery charging and discharging. The microcontroller intelligently controls the transition between operating modes by continuously assessing power demand and availability to identify the best source. A key component of the system is the Adaptive Spatio-Temporal Graph Convolutional Learning (ASTGCL) model, which enables proactive and dynamic energy management by forecasting the best power dispatch based on environmental data, resource availability, and load demand. Real-time monitoring, performance tracking, and remote control for increased efficiency are made possible with IoT integration. Real-time hardware implementation and simulation studies are used to validate the proposed model. The simulation results show that the proposed technique outperforms conventional models in terms of achieving quick, seamless switching, guaranteeing a continuous supply and maximizing the usage of renewable energy. With a mode-switching time of only 24 ms, the proposed MCU-based EMS significantly outperforms other conventional methods like MGIB (120 ms), NILM (95 ms), M-GRU-based NILM (110 ms) and FF-SHMS (105 ms).