A Time-aware Method for Occupancy Detection in a Building

The target of buildings’ energy efficient is to facilitate a comfortable environment for occupants while maintaining minimal energy consumption. Occupant behaviors pay a large impact in influencing the energy consumption. Time-aware occupancy detection could give information support for intelligent building energy management. In this paper several building occupancy detection methods, which are based on the temporal analysis of historical data, are proposed for providing different size of prediction window occupancy detection. Each proposed approaches are evaluated against accurate real-life data collected from a building. Experiments have been conducted using actual occupancy data under six different time horizons can be used to perform time-aware occupancy states prediction. The experimental results show that Stochastic Gradient Descent (SGD) and Gaussian mixture models-Hidden Markov Model (GMM-HMM) outperforms the other methods for the evaluation. With proposed more accurate time-aware occupancy prediction algorithms, we hope to develop more energy-efficient HVAC(Heating, Ventilation, and Air Conditioning) scheduling systems in order to save energy consumption.