Automatic Real-Time Prediction of Energy Consumption Based on Occupancy Pattern for Energy Efficiency Management in Buildings

Abstract

Energy usage of non-domestic buildings in the UK accounts for a significant portion of total energy consumption and CO2 emissions. Occupant behaviour and comfort management have a significant impact on the total energy consumption in buildings. However, current building energy management systems generally operate based on fixed schedule, maximum design occupancy assumption and pre-defined occupant comfort levels to ensure satisfactory temperatures, ventilations and luminance at all times by manual configuration beforehand. This is costly and inefficient. In this work, we have proposed an automated approach based on probabilistic machine learning to model and predict energy consumption using occupancy data for energy efficiency management in non-domestic buildings. The proposed approach is able to predict energy consumption and detect anomaly energy usage in real time. It has been validated with real datasets collected from a non-domestic building. The experimental results have demonstrated the effectiveness of the proposed system.