An Anomaly Detection Model for Enhancing Energy Management in Smart Buildings

Smart buildings provide an excellent opportunity to monitor the energy consumption behavior. It can assist the building management to find unexpected energy usage patterns. In this research, we present our model to find abnormal energy consumption patterns by analyzing the temporal data streams gathered from smart meters. We investigate support vector regression with radial basis function to find the mismatch between actual and expected energy consumption. It has the ability to map the non-linearity of data and predict expected energy consumption. We build the energy usage profile and provide visualization services over it. Furthermore, energy profiles may be used for different objectives including customer classification and load forecasting. In this preliminary study, we performed the experiments over a real electrical load measurements dataset collected from a dwelling. The obtained results suggest that our proposed model is feasible and practical solution to detect anomalies and provide good insight to visualize the energy consumption behavior.