Smart microgrid management based on predictive control and demand forecasting

Efficient management of microgrids in modern electrical systems allows for optimized use of energy resources and improved integration into the electrical system. This study presents an approach that combines predictive control with demand forecasting to anticipate variations in consumption and generation based on renewable resources such as solar, wind, and hydrokinetic energy. The predictive control model incorporates robust control based on the H_∞ theory combined with the Wide Neural Network (WNN) model for demand prediction. This design provides robust SOC regulation of the storage system and accurate power balancing with disturbances in the range of ±7 kW. Data acquisition reflects its application in real-world environments in generation and demand, enabling improved adaptation in smart, operational, and adaptive microgrids. Demand prediction results were evaluated using RMSE, S-Square, MSE, and MAPE indicators. The WNN model stands out for the database studied at the research focus, with an RMSE of 0.68045 and a MAPE of 6.4%. Finally, it offers a substantial improvement in energy balance, mitigates battery strain, and ensures resilient operation for the next generation of smart grids.