Enabling real-world deployment of data driven pre-cooling in smart buildings

摘要

Facility managers of commercial buildings are confronting a challenging problem, that of reducing the peak demand, energy consumption and operating energy bills of their buildings. The integration of Internet of Things (IoT) in buildings is opening the doors for novel data driven techniques to address these challenges. In this paper, we focus on one such technique, called pre-cooling, and demonstrate its practical potential for energy and cost savings by applying it to a large office building located in Australia. Our contributions are threefold. First, to enable real-world deployment of pre-cooling, we make a case for why it is critical to forecast two key quantities, namely outside air (ambient) temperature and occupancy. Second, we develop mechanisms to forecast these quantities and show that the internal zone temperature predicted by our model in a day-ahead manner matches very well with the actual zone temperature measurements. The root mean square error (RMSE) is low, around 0.10° C on average. Finally, we feed the forecasts into our pre-cooling energy-cost optimization framework and quantify the performance across several days in the high-demand summer time-frame of January 2017. The results point to substantial savings - peak power can be reduced by up to 35%, energy consumption by up to 28% and energy bills by up to 34%. Our optimal pre-cooling solution is ready for real-world deployment and empowers facility managers with a low cost solution for improving the energy and cost footprints of their buildings.