An approach for building occupancy modelling considering the urban context
Introduction
Building occupancy is a key parameter of building energy modelling and a basic factor of modelling occupant behaviour [1]. Bahaj and James [2] discovered that different occupancy patterns lead to different energy consumptions in identical buildings. Therefore, accurately modelling and comprehensively understanding how people use building can help improve energy efficiency [3]. Although the IEA Annex 66 on Definition and Simulation of Occupant Behaviour in Buildings has given guidance on the current state of occupancy modelling, there are still unanswered questions relating to the enhancement of modelling techniques, from the sensing, interpretation and prediction perspectives [4].
Driven by multidisciplinary factors, occupancy has three features that pose challenges to occupancy modelling: stochasticity, diversity and complexity [4]. Stochasticity has been the research focus over a past decade and has enriched the methodological approaches for modelling random occupancy chain. In fact, occupancy is not purely random but could also be characterised by deterministic factors in surrounding systems. These factors could be the internal conditions of building environment and/or the conditions external to the building system. Such conditions can be natural (weather and local topology) as well as anthropogenic, related to urban systems (conditions of transport network and presence of event). Incorporating both of those conditions as potential driving factors of building occupancy into a modelling framework could be a promising way of understanding and enhancing the way we model the complexity and diversity of occupancy patterns.
Considering spatial and temporal dependencies between building occupancy and conditions on surrounding urban systems, this paper proposes a modelling framework that comprehensively and flexibly accounts for such interactions. The framework is operationalised by the means of a competing risk hazard model using implicit occupancy data inferred from wireless (Wi-Fi) networks in the case study area. This approach has apparently not been applied in this discipline so far and has capabilities exceeding those of the more conventional approaches, e.g. MCMs. In particular, it can more easily incorporate multiple exogenous variables simultaneously and allow insights into exogenous impacts on occupancy and thus energy consumption. As an additional contribution, this paper also presents and applies a methodology of processing customarily available Wi-Fi data for occupancy modelling.
Section snippets
Literature review
The review is structured to correspond to the three contributions of this paper. Section 2.1 summarises studies related to interactions between human behaviour and factors originating in urban systems. Section 2.2 comparatively discusses the applications of MCMs and hazard-based models in occupancy modelling to date. Section 2.3 reviews different approaches to implicit occupancy detection, including discussion of the challenges associated with the use of Wi-Fi data for occupancy modelling in
Methodology
The following sections describe the urban-system level modelling framework, the HBM approach, the development of the conventional discrete-time MCM, the case study description, Wi-Fi data pre-processing and the performance metrics used to evaluate the two methods (HBM and MCM) from the estimation, interpretation and prediction perspectives.
Comparison of transition probability distributions
Comparison between the MCM and HBM with respect to the distribution of the transition probabilities requires plotting of the underlying PDFs. However, a PDF underlying a HBM parameterised using covariates will also depend on those covariates, themselves being particular to the circumstances they reflect. Hence for the purpose of demonstrating how the HBM captures more complex duration-dependencies in the transition distributions when compared to the MCM, we use the example of a baseline HBM
Conclusions
In this paper, we propose an urban-system level framework conceptualising the interactions between building occupancy and urban systems. We demonstrate how the framework can be operationalised using a competing risk hazard model, incorporating Wi-Fi connection logs, to analyse the occupancy of Imperial College Faculty Building in April–May 2017. A time-inhomogeneous discrete-time MCM was also implemented as the reference method. Comparing the performance of the two models, we identify three key
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This research was financially supported by the China Scholarship Council and Imperial College London. The authors would like to appreciate Mr. Matthew Balyuzi (Networks Project Team Leader, Imperial College London) for providing Wi-Fi connection data and discussing associated protocol issues. The authors acknowledge the support from the Integrated Development of Low-Carbon Energy Systems (IDLES) research programme at Imperial College London funded by the Engineering and Physical Sciences
References (65)
- et al.
Simulation of occupancy in buildings
Energy Build.
(2015) - et al.
Urban energy generation: the added value of photovoltaics in social housing
Renew. Sustain. Energy Rev.
(2007) - et al.
IEA EBC Annex 66: Definition and simulation of occupant behavior in buildings
Energy Build.
(2017) - et al.
An ontology to represent energy-related occupant behavior in buildings. Part I: introduction to the DNAs framework
Build. Environ.
(2015) - et al.
An ontology to represent energy-related occupant behavior in buildings. Part II: implementation of the DNAS framework using an XML schema
Build. Environ.
(2015) - et al.
A generalised stochastic model for the simulation of occupant presence
Energy Build.
(2008) - et al.
Modeling regular occupancy in commercial buildings using stochastic models
Energy Build.
(2015) - et al.
A context-aware method for building occupancy prediction
Energy Build.
(2016) - et al.
Occupancy prediction through Markov based feedback recurrent neural network (M-FRNN) algorithm with WiFi probe technology
Build. Environ.
(2018) - et al.
A bottom-up stochastic model to predict building occupants' time-dependent activities
Build. Environ.
(2013)