A Probabilistic Model to Predict Household Occupancy Profiles for Home Energy Management Applications

Due to the impact of human lifestyle on building energy consumption, the development of occupants' behavior models is crucial for energy-saving purposes. In this regard, occupancy modeling is an effective approach to intend such a purpose. However, the literature reveals that existing occupancy...

Full description

Saved in:
Bibliographic Details
Published in:IEEE access 2021, Vol.9, p.38187-38201
Main Authors: Rueda, Luis, Sansregret, Simon, Le Lostec, Brice, Agbossou, Kodjo, Henao, Nilson, Kelouwani, Sousso
Format: Article
Language:English
Subjects:
behavior
Buildings
Cox regression
Energy consumption
Energy management
hazard rate
Hidden Markov models
Human behavior
Load modeling
Markov chains
Markov processes
markov-chain
Occupancy
Parametric analysis
Predictive models
Probabilistic logic
Probabilistic models
Regression analysis
Regression models
Residential buildings
Residential energy
Statistical analysis
Statistical models
survival analysis
Variability
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Due to the impact of human lifestyle on building energy consumption, the development of occupants' behavior models is crucial for energy-saving purposes. In this regard, occupancy modeling is an effective approach to intend such a purpose. However, the literature reveals that existing occupancy models have limitations related to the representation of occupancy state duration and the integration of occupancy variability among individuals. Accordingly, this paper proposes an explicit differentiated duration probabilistic model to generate realistic daily occupancy profiles in residential buildings. The discrete-time Markov chain theory and the semi-parametric Cox proportional hazards model (Cox regression) are used to predict household occupancy profiles. The proposed model is able to capture occupancy states duration and integrate human behavior variability according to individuals' characteristics. Moreover, a parametric analysis is employed to investigate these characteristics' impact on the model performance and consequently, select the most significant input variables. A validation process is conducted by comparing the model performance with that of previous methods, presented in the literature. For this purpose, the k crossvalidation technique is utilized. Validation results demonstrate that the proposed approach is highly efficient in generating realistic household occupancy profiles.
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2021.3063502