<  Back to the Polytechnique Montréal portal

Autoregressive neural networks with exogenous variables for indoor temperature prediction in buildings

Benoit Delcroix, Jérôme Le Ny, Michel Bernier, Muhammad Azam, Bingrui Qu and Jean-Simon Venne

Article (2020)

[img] Accepted Version
Restricted to: Repository staff only until 21 February 2021.
Request a copy
Cite this document: Delcroix, B., Le Ny, J., Bernier, M., Azam, M., Qu, B. & Venne, J.-S. (2020). Autoregressive neural networks with exogenous variables for indoor temperature prediction in buildings. Building Simulation. doi:10.1007/s12273-019-0597-2
Show abstract Hide abstract

Abstract

Thermal models of buildings are helpful to forecast their energy use and to enhance the control of their mechanical systems. However, these models are building-specific and require a tedious, error-prone and time-consuming development effort relying on skilled building energy modelers. Compared to white-box and gray-box models, data-driven (black-box) models require less development time and a minimal amount of information about the building characteristics. In this paper, autoregressive neural network models are compared to gray-box and black-box linear models to simulate indoor temperatures. These models are trained, validated and compared to actual experimental data obtained for an existing commercial building in Montreal (QC, Canada) equipped with roof top units for air conditioning. Results show that neural networks mimic more accurately the thermal behavior of the building when limited information is available, compared to gray-box and black-box linear models. The gray-box model does not perform adequately due to its under-parameterized nature, while the linear models cannot capture non-linear phenomena such as radiative heat transfer and occupancy. Therefore, the neural network models outperform the alternative models in the presented application, reaching a coefficient of determination R2 up to 0.824 and a root mean square error down to 1.11 °C, including the error propagation over time for a 1-week period with a 5-minute time-step. When considering a 50-hour time horizon, the best neural networks reach a much lower root mean square error of around 0.6 °C, which is suitable for applications such as model predictive control.

Open Access document in PolyPublie
Subjects: 2100 Génie mécanique > 2100 Génie mécanique
2100 Génie mécanique > 2109 Instrumentation et systèmes mécaniques
2500 Génie électrique et électronique > 2500 Génie électrique et électronique
2600 Robotique > 2604 Applications de systèmes intelligents
Department: Département de génie électrique
Département de génie mécanique
Research Center: Non applicable
Funders: IVADO
Date Deposited: 02 Mar 2020 11:26
Last Modified: 03 Mar 2020 01:20
PolyPublie URL: https://publications.polymtl.ca/4220/
Document issued by the official publisher
Journal Title: Building Simulation
Publisher: Springer
Official URL: https://doi.org/10.1007/s12273-019-0597-2

Statistics

Total downloads

Downloads per month in the last year

Origin of downloads

Dimensions

Repository Staff Only