Skip to Main content Skip to Navigation

Commande prédictive hiérarchisée hybride pour la gestion de l’énergie dans les bâtiments

Abstract : Intelligent management strategies to optimize building energy consumption are considerably gaining attention due to the current climate challenges and the technological evolution of the automation solutions. To tackle the new energy efficiency standards, building energy management systems must be able to control energy consuming devices in order to minimize the costs and optimize the comfort of occupants. In this thesis, we study hybrid multitime scale model predictive control strategies to tackle building energy management problems. We proposed a two-layer hierarchical controller to jointly control the energy consumption and power demand of the system. The upper level implements a long term economic optimization that takes into account the energy price and the requirements of the occupants. The lower layer ensures the tracking of the optimal scheduling plan computed by the upper layer with a shorter prediction horizon and a higher sampling rate. Two topics related to the operation of multi-scale controllers are considered. The first is the interaction between the levels of optimization and the consistency of the information exchanged. We investigate different strategies to project the results of the upper layer at the lower one and provide comparisons to highlight its impact on the closed-loop behavior. The second is the management of On/Off loads in the multi-time scale framework. A geometric analysis of the decision space of the long-term optimization problem is performed to study the consequences of adding binary constraints to the problem. Then, we propose a reformulation strategy to improve the quality of the final control and limit the effort required to find the solution.
Complete list of metadatas

Cited literature [123 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Wednesday, February 26, 2020 - 4:17:33 PM
Last modification on : Thursday, March 12, 2020 - 4:20:35 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02492087, version 1


Amanda Abreu de Oliveira. Commande prédictive hiérarchisée hybride pour la gestion de l’énergie dans les bâtiments. Automatique / Robotique. CentraleSupélec, 2019. Français. ⟨NNT : 2019CSUP0004⟩. ⟨tel-02492087⟩



Record views


Files downloads