Abstract

The rapid expansion of smart home technologies has enabled high-resolution monitoring of appliance-level electricity consumption. This study presents a hybrid framework that combines machine-learning-based prediction with discrete-event simulation to detect and analyze periods of elevated energy usage in smart residential environments. A Random Forest classifier was trained using appliance-level time-stamped wattage readings enriched with temporal and behavioral features. Time-series-aware cross-validation was employed to prevent data leakage and ensure realistic model performance.

To evaluate real-time applicability, the trained classifier was integrated into a SimPy-based simulation that generates synthetic appliance behavior over an extended weekly horizon. Experimental results show that the model achieves high predictive performance (accuracy β‰ˆ 0.986) and maintains stable behavior when deployed within a dynamic simulated environment. The proposed framework provides a practical foundation for energy optimization, demand-response strategies, and intelligent automation systems in smart homes.

Keywords

  • Energy Prediction
  • Smart Home Analytics
  • Random Forest
  • Simulation
  • SimPy
  • High Consumption Detection

References

  1. 1. Zeifman, M., & Roth, K. (2011). Nonintrusive appliance load monitoring: Review and outlook. IEEE Transactions on Consumer Electronics, 57(1), 76–84.
  2. 2. Liaw, A., & Wiener, M. (2002). Classification and regression by randomForest. R News, 2(3), 18–22.
  3. 3. SimPy Documentation. (2022). SimPy: Discrete-event simulation for Python. Available online: https://simpy.readthedocs.io
  4. 4. Zhang, C.; & Li, F. (2015). Smart home energy management system using machine learning. Energy Procedia, 75, 174–179.
  5. 5. Arrubla-Hoyos, W.; & Severiche Maury, Z. (2025). Energy Consumption Dataset For Smart Homes [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.14768659