OCCUPANT BEHAVIOR AND SCHEDULE PREDICTION BASED ON OFFICE APPLIANCE ENERGY CONSUMPTION DATA MINING

Plug load and Heating Ventilation and Air Conditioning (HVAC) systems are the two largest energy consumers in commercial buildings. The use of the 2 systems is closely related to occupant behavior and schedule. It is significant to learn the occupant behavior pattern and predict occupancy schedule for controlling the systems to save energy. A data mining study is performed on the office appliance energy consumption data to predict the individual occupant behavior and the occupancy schedule in an open office space. An experiment is conducted for 2 weeks by using wireless electric outlet meters. The 5-minute interval electricity consumption data of computers, computer monitors, task lights, and other office appliances are monitored for 6 office workers. Occupant behavior is categorized as “Occupied computer-based work”, “Occupied non-computer-based work”, “Unoccupied remote work”, and “Unoccupied”. C4.5 algorithm is used for pattern recognition over the appliance electricity consumption data individually. The average percentage of correct of the 6 individuals is 92.39 % using 10-fold cross validation. The occupancy schedule for the space is predicted by using total energy consumption of each subject with Linear Regression algorithm. The correlation coefficient is 0.92 using 10-fold cross validation. The results suggest the models are feasible and can be applied to the plug load and HVAC control systems to reduce energy consumptions