۱- Introduction

۲- Method

۳- Results

۴- Discussion

۵- Limitations

۶- Conclusion

References

Abstract

This paper addresses the question of identifying the time-window in short-term past from which the information regarding the future occupant’s window opening actions and resulting window states in buildings can be predicted. The addressed sequence duration was in the range between 30 and 240 time-steps of indoor climate data, where the applied temporal discretization was one minute. For that purpose, a deep neural network is trained to predict the window states, where the input sequence duration is handled as an additional hyperparameter. Eventually, the relationship between the prediction accuracy and the time lag of the predicted window state in future is analyzed. The results pointed out, that the optimal predictive performance was achieved for the case where 60 time-steps of the indoor climate data were used as input. Additionally, the results showed that very long sequences (120–۲۴۰ time-steps) could be addressed efficiently, given the right hyperprameters. Hence, the use of the memory over previous hours of high resolution indoor climate data did not improve the predictive performance, when compared to the case where 30/60 min indoor sequences were used. The analysis of the prediction accuracy in the form of F1 score for the different time lag of future window states dropped from 0.51 to 0.27, when shifting the prediction target from 10 to 60 min in future.

Introduction

Occupant behavior (OB) has been identified to be one of the principal factors influencing the energy consumption in commercial buildings [1], [2], [3], [4]. Due to that, developing an accurate model that predicts human actions would be beneficial for achieving higher indoor comfort or optimization of energy consumption. Additionally, there have been a number of studies that addressed modeling the OB for its inclusion in building automation systems (BAS) [5], [6], [7], [8], [9], [10], [11]. According to the current research, OB in buildings is often defined as a discrete sequence in the temporal domain [5], [12], [13], [14], [15], [16], [17], [18]. As such, not only is it necessary to identify which variables lead to occupants’ actions, but also in which temporal rangedo the changes of variables in question occur, which motivated a number of studies on time-series modeling of OB. Liao et al. [14] presented a probabilistic graphical model for depicting the time-series of occupancy data. Fritsch et al. [12] presented the model that generates time series of window opening angles with the same statistics as the measured openings for the heating period. Dong and Andrews [5] proposed occupancy pattern recognition using semi-Markov models. Youngbloot and Cook [13] introduced a hierarchical model for controlling the smart environment based on occupants’ activities and concluded that learning algorithms built on Markov models experience performance issues when scaled to large problems.