Predictive Monitoring of Energy Efficiency: Supervised Learning Case Study

Abstract

Rising energy consumption in buildings and its environmental impact have brought significant attention to energy efficiency and monitoring practices. Many institutions adopt energy management systems aligned with ISO 50001, which emphasizes continuous monitoring and improvement. In this context, artificial intelligence and machine learning offer effective tools for detecting anomalies and forecasting energy demand. This study presents a practical approach using supervised learning models-Random Forest and Long Short-Term Memory (LSTM) networks-for short-term consumption forecasting and anomaly detection. The methodology relies on real-world data from a high-consumption university building, including manually labeled anomalies and high-frequency electrical measurements. Results show that both models are robust under noisy conditions, adaptable to different sampling resolutions, and benefit from periodic retraining. These findings support their integration into smart monitoring platforms aligned with ISO 50001 goals.