Enabling building energy auditing using adapted occupancy models

Understanding building energy consumption has become important due to stricter energy regulations, increasing energy costs and also as buildings have long term impact on energy consumption. In order to recommend retrofits, it is important to have accurate estimates for building energy consumption which is affected significantly by occupancy patterns. This paper explores the development of static occupancy models using a model adaptation technique that is able to capture accurately features of occupancy distributions typically found using a large amount of training data (days, weeks, months). Using only one day of training data that can be easily recorded without any infrastructure but battery-operated sensors with on-board memory, we show that our adapted occupancy model can estimate energy savings of 10.9%; and the room temperatures for the adapted model schedules were 0.5°F and 1.4°F off from the target temperatures for summer and winter months, respectively. This performance was on par with models trained with four times as much data. Our proposed technique can be used by energy auditors to estimate energy savings for existing buildings and by building energy managers to optimize static schedules which assume maximum occupancy.