Ph. D. Project
Title:
Intelligent data-driven and occupant-centric control for building energy efficiency
Dates:
2023/05/23 - 2026/05/24
Student:
Supervisor(s):
Other supervisor(s):
Prof. JAMOULI Hicham (h.jamouli@uiz.ac.ma)
Description:
The proposed thesis project is part of the active energy efficiency of buildings with the objective of developing new approaches to modeling, simulation and control of energy efficiency of buildings in order to reduce their consumption in operational phase. A lot of work has been done in the last few years on this active approach and several innovative solutions to achieve energy efficiency objectives in the presence of various constraints have been proposed mainly in the context of energy management and indoor comfort control systems [1]. Advanced automatic control techniques and in particular predictive control based on constrained optimization techniques have led to significant progress in building energy management. These techniques have been used both for efficient energy management in nominal operation and in faulty equipment situations to accommodate faults and reduce energy consumption when fault modes occur [2].
To significantly improve the energy efficiency of buildings in which people live and/or work, spending on average 65% to 90% of their time, the scientific literature shows that it will not be enough to modify construction techniques or to use more energy efficient control technologies . Among the many factors on which the energy consumption of a building depends (e.g., envelope, orientation, and environmental characteristics such as heating, cooling, lighting, etc.), it turns out that the occupant and his or her behavior are extremely important key factors that shape and strongly influence energy use in complex and sometimes counterintuitive ways [3]. Indeed, occupant actions, such as setting a thermostat for comfort, turning lights on and off, using appliances, opening and closing windows, raising and lowering blinds and moving between spaces, etc., can have a significant impact on actual energy consumption and indoor comfort. In the operational phase of buildings, works taking into account the key non-technical factor of human behavior in the issue of active control of energy efficiency are indeed rare, except for some works reported very recently in the scientific literature [4].
In view of this gap in research on active control of energy efficiency focused on the occupant and his behavior, the scientific objectives of this thesis project are to develop dynamic models of buildings integrating the behavior of the occupants and oriented towards the control with the aim of designing and implementing control strategies for an optimal reduction of energy consumption. The original and innovative point of view that we wish to adopt in this research work is to consider the occupants' behavior as an endogenous characteristic of the global dynamics of a building. An important research question that follows from this point of view is how this characteristic can be correctly identified and modeled for an intelligent control of energy efficiency? The research project proposes to break down this important question into two central issues:
i. dynamic building modeling focused on occupant behavior and adapted to the objective of optimal energy efficiency control
ii. the development of new techniques and algorithms for occupant-centered adaptive intelligent control, which ensure the energy efficiency of buildings under the constraint of thermal comfort and human behavior variations.
This question of intelligent control of the energy efficiency of buildings as formulated via the above problems presents specific non-trivial difficulties and scientific locks to be removed because of the cumulative constraints that they imply with regard to thermal modeling and the taking into account of occupancy and occupant behavior.
A research route that we plan to explore at the methodological level is that of modeling and control based on data that is now omnipresent thanks to digital technologies. The analysis of collected data can help to understand the way occupants react to their indoor environment and consequently help to reveal certain parameters related to occupant information. The identification of these parameters (e.g., occupancy, location, behavior/activity, movement, etc.) will therefore be crucial to "inject" intelligence into energy efficiency control systems. The envisaged control strategies, centered on the occupant, will focus on data-driven strategies with a particular emphasis on optimal control by reinforcement learning and its possible variants of predictive control based on reinforcement learning. The feasibility of the results will be demonstrated via numerical simulations in the MATLAB environment coupled if necessary with specific building simulation software (e.g., TRNSYS) and via the CRAN's newly built Eco-safe demonstrator installed at ATELA.
To significantly improve the energy efficiency of buildings in which people live and/or work, spending on average 65% to 90% of their time, the scientific literature shows that it will not be enough to modify construction techniques or to use more energy efficient control technologies . Among the many factors on which the energy consumption of a building depends (e.g., envelope, orientation, and environmental characteristics such as heating, cooling, lighting, etc.), it turns out that the occupant and his or her behavior are extremely important key factors that shape and strongly influence energy use in complex and sometimes counterintuitive ways [3]. Indeed, occupant actions, such as setting a thermostat for comfort, turning lights on and off, using appliances, opening and closing windows, raising and lowering blinds and moving between spaces, etc., can have a significant impact on actual energy consumption and indoor comfort. In the operational phase of buildings, works taking into account the key non-technical factor of human behavior in the issue of active control of energy efficiency are indeed rare, except for some works reported very recently in the scientific literature [4].
In view of this gap in research on active control of energy efficiency focused on the occupant and his behavior, the scientific objectives of this thesis project are to develop dynamic models of buildings integrating the behavior of the occupants and oriented towards the control with the aim of designing and implementing control strategies for an optimal reduction of energy consumption. The original and innovative point of view that we wish to adopt in this research work is to consider the occupants' behavior as an endogenous characteristic of the global dynamics of a building. An important research question that follows from this point of view is how this characteristic can be correctly identified and modeled for an intelligent control of energy efficiency? The research project proposes to break down this important question into two central issues:
i. dynamic building modeling focused on occupant behavior and adapted to the objective of optimal energy efficiency control
ii. the development of new techniques and algorithms for occupant-centered adaptive intelligent control, which ensure the energy efficiency of buildings under the constraint of thermal comfort and human behavior variations.
This question of intelligent control of the energy efficiency of buildings as formulated via the above problems presents specific non-trivial difficulties and scientific locks to be removed because of the cumulative constraints that they imply with regard to thermal modeling and the taking into account of occupancy and occupant behavior.
A research route that we plan to explore at the methodological level is that of modeling and control based on data that is now omnipresent thanks to digital technologies. The analysis of collected data can help to understand the way occupants react to their indoor environment and consequently help to reveal certain parameters related to occupant information. The identification of these parameters (e.g., occupancy, location, behavior/activity, movement, etc.) will therefore be crucial to "inject" intelligence into energy efficiency control systems. The envisaged control strategies, centered on the occupant, will focus on data-driven strategies with a particular emphasis on optimal control by reinforcement learning and its possible variants of predictive control based on reinforcement learning. The feasibility of the results will be demonstrated via numerical simulations in the MATLAB environment coupled if necessary with specific building simulation software (e.g., TRNSYS) and via the CRAN's newly built Eco-safe demonstrator installed at ATELA.
Department(s):
| Control Identification Diagnosis |
