Smart Building Energy Management System Using IoT and Artificial Intelligence

Abstract

Buildings represent a major share of electricity consumption, and their operational energy use is strongly influenced by weather, occupancy dynamics, and the interaction of heating, ventilation, and air-conditioning systems with lighting and plug loads. Conventional building operation typically relies on fixed schedules and manually tuned control rules, which often fail to adapt to daily variability and can lead to unnecessary energy use and peak-demand events. This paper proposes a Smart Building Energy Management System using Internet of Things sensing and artificial intelligence to improve energy efficiency while maintaining indoor comfort and operational robustness. The objective of the study was to design an end-to-end, deployable pipeline that integrates sensing, data processing, load forecasting, and control optimization in a closed-loop framework. An applied experimental prototyping approach was used. Time-series data were collected from the system operation, synchronized, cleaned, and transformed through feature engineering. Two forecasting models were evaluated for short-term demand prediction: a gradient-boosting model using engineered features and a recurrent neural network model using temporal sequences. The control strategy was implemented in stages, starting from a safe rule-based baseline and extending to forecast-informed bounded optimization with fallback mechanisms to ensure reliable operation. The results showed that the proposed system reduced total energy consumption by 13.3% and reduced maximum peak demand by 11.5% during a 28-day evaluation period. Comfort compliance improved, as indicated by a 19.0% reduction in total comfort-violation duration. The recurrent neural network model achieved lower forecasting error than the gradient-boosting model across the evaluated metrics. In conclusion, integrating connected sensing with artificial intelligence-based forecasting and constraint-aware optimization can deliver measurable energy and demand reductions while maintaining comfort. The study contributes an implementable methodology and reporting structure for smart building energy management, while future work should validate performance in longer deployments and diverse building types under seasonal and occupancy variation.