Skip to Main content Skip to Navigation
Theses

Amélioration de la prédiction de quantités d'intérêt par modélisation inverse : application à la thermique du bâtiment

Abstract : This work introduces an original inverse strategy for model parameter identification that can be used for onsite building characterization in view of energy performance assessment and as a tool of decision-making during energy retrofitting of existing buildings. Unlike the standard global inverse approaches such as Tikhonov regularization method that aim at identifying all the model parameters in order to best fit the measurement data, the goal-oriented inverse method is formulated for a robust prediction of a quantity of interest. Thus, it only updates the model parameters that most affect the computation of the sought quantity of interest. In order to optimize the computation time, the goal-oriented inverse method is combined with the Proper Generalized Decomposition (PGD) model order reduction method. The proposed identification strategy is applied to two existing buildings part of the equipment “Sense-City” that were instrumented for this purpose. The results show that the goal-oriented inverse method robustly predicts the sought quantities of interest by only updating the model parameters to which they are sensitive and it converges faster than the Tikhonov regularization method. Finally, the proposed inverse strategy can be applied to occupied buildings and extended to the district scale. It can also be used for the optimal placement of sensors
Complete list of metadatas

Cited literature [169 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02477069
Contributor : Abes Star :  Contact
Submitted on : Thursday, February 13, 2020 - 10:35:17 AM
Last modification on : Friday, July 17, 2020 - 5:08:53 PM
Long-term archiving on: : Thursday, May 14, 2020 - 1:20:47 PM

File

TH2019PESC2006.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02477069, version 1

Collections

Citation

Zohra Djatouti. Amélioration de la prédiction de quantités d'intérêt par modélisation inverse : application à la thermique du bâtiment. Thermique [physics.class-ph]. Université Paris-Est, 2019. Français. ⟨NNT : 2019PESC2006⟩. ⟨tel-02477069⟩

Share

Metrics

Record views

121

Files downloads

84