An occupancy prediction model for campus buildings based on the diversity of occupancy patterns

Abstract

The complexity and randomness of occupancy often lead to the deviation between simulated energy and measured energy. It is of great significance to select appropriate approaches for occupancy prediction. This study applied Gaussian distribution model to fit occupancy of three functional buildings within a campus. Abandoned the average occupancy level widely adopted in exiting researches, a novel classified approach considering the diversity of occupancy patterns was proposed. The resulting Gaussian curves presented a better fitting performance for stable changes of occupancy. Although the sudden increase and decrease of occupancy greatly affects the prediction accuracy, the occupancy prediction error based on Gaussian distribution can still be controlled within ±15 %. The energy of the case building was obtained by superposing simulated energy of each occupancy pattern. Rather than traditional method of averaging, the classified methodology eliminated the simulation errors produced from the inhomogeneous and stochastic occupancy. The exploration of changing regularity of energy affected by occupancy and time periods with energy saving potential were achieved by integrating energy data with occupancy data. And the detection of the degree and occurrence times of peak occupancy of various occupancy patterns of rooms provided a support for rational load distribution.