Anomaly Detection in building energy system using machine learning technique

Résumé

Background Energy consumption in buildings has steadily increased over time due to growing urbanization and the need for living more comfortably and working environments. The growing inefficiencies in energy consumption within buildings are caused by reliance on static monitoring systems, which often failed to detect subtle or evolving anomalies. The inability to adapt to dynamic consumption patterns, resulting in waste, higher costs, and reduced sustainability objective was to design a model capable of learning from data, identifying abnormal energy usage, and dynamically adjusting to new conditions. Methods Support Vector Machine (SVM) was applied using electricity consumption data sourced from smart meters. The dataset underwent cleaning, feature extraction, and scaling before model training and testing. The developed model achieved reliable anomaly detection, identifying irregular energy consumption with improved accuracy compared to traditional rule-based methods. Results The performance metrics of the model was Accuracy–0.97, F1 Score-0.73, Precision-0.79, Recall-0.68. The results showed reduced false alarms and enhanced adaptability to changes in energy use patterns. Conclusion The study contributed to energy efficiency, cost reduction, and sustainable building management, while also providing a replicable framework for intelligent energy monitoring. The benefits extended to facility managers and researchers by offering an adaptable system for smarter energy decision-making.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0