Buildings account for a large portion of the total energy consumption and they might serve as a significant thermal storage capacity that can be advantageous for the future energy grid. To utilise this capacity, it is necessary to characterise the thermal dynamics in buildings using methods that are general enough to be applicable to a significant share of the building stock. This work proposes a data-driven method to characterise thermal dynamics of thermostatically controlled buildings with night setback. The method includes 1) using Hidden Markov Models to systematically select data periods when the indoor temperature decays steadily during night; 2) model reduction of a Stochastic Differential Equations model of heat transfer to a discrete linear model which is fitted by utilising the selected night-time data; and 3) computing one short time constant and one long time constant, which allows to categorise buildings according to their thermal response. This method is applied to 39 different Danish residential buildings and the results reveal that this simplified model captures the main processes governing the heat transfer: the one-step predictions for the indoor air temperature return 95% of the residuals . For all buildings, the short time constants are lower than an hour, and the long time constants range from 20 h to 100 h. Finally, this method is used in simulated data to validate that the time constants provide insight about the energy flexibility potential of a building. The results show that dynamic thermal response of buildings can be discovered using limited data.
Bibliografisk noteFunding Information:
This work is part of the CITIES project (nr. DSF1305-00027B) as well as the Smart Energi i Hjemmet project. I would like to thank my colleague Rune G. Junker for many valuable discussions and for sharing his work and vision on the future of flexible energy systems.
Copyright 2020 Elsevier B.V., All rights reserved.
- Data analysis
- Demand response
- Energy flexibility
- Thermal building characterisation