Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Modeling and Control of Energy Consumption in Buildings: Thermal and Visual Comfort
In the context of global warming, a challenging issue is reducing the amount of energy consumption, specifically in the building sector and meanwhile respecting the existing constraints. This necessitates utilizing advanced modelling techniques and control strategies for buildings energy systems.
Keywords: Energy, Building, Machine Learning, System Identification, MPC
The energy consumed in the buildings is mainly according to heating, cooling, lighting and air conditioning. In the heating seasons, while the solar radiation is a source of free energy, it can induce visual discomfort. Similarly, in the cooling seasons like summer, the solar radiation is a disturbance to be rejected using blinds. This can also induce illuminance reduction and subsequently, visual discomfort. Accordingly, one can see that minimizing the energy consumption or the monetary costs and meanwhile satisfying the thermal and visual comfort is a complicated task, which demands advance control methods. For applying such methods like model predictive control (MPC), it is necessary to have a mathematical model for the energy and lighting system in the building. This includes the model for the thermal dynamics of building besides the mathematical representations of visual comfort (DGP and workplace illuminance) as a functions of state, input and disturbance variables of system. For obtaining these models, we need informative measurement data collected by performing specific suitable experiments. These experiments should be carefully designed and implemented. Once the data is available, one can use system identification and machine learning techniques in order to obtain suitable mathematical description of the system and other essential models. For implementation of MPC, we need also the prediction of climate variables and possibly the prices of electrical power and thermal energies. Toward the formulation of control problem, it is also necessary to define a cost function that reflects the energy consumption, the monetary cost, and/or the discomfort of users.
Various issues can affect the performance of the implemented MPC such as the sampling time and horizon of MPC. However, the most important factor is the quality of derived mathematical models, i.e., if the models are not reasonably precise in predicting, the performance of MPC will not be satisfactory. Therefore, a main task in this approach is obtaining high quality models, verifying them in simulation and practice, and finally optimizing the general performance by tuning parameters.
The energy consumed in the buildings is mainly according to heating, cooling, lighting and air conditioning. In the heating seasons, while the solar radiation is a source of free energy, it can induce visual discomfort. Similarly, in the cooling seasons like summer, the solar radiation is a disturbance to be rejected using blinds. This can also induce illuminance reduction and subsequently, visual discomfort. Accordingly, one can see that minimizing the energy consumption or the monetary costs and meanwhile satisfying the thermal and visual comfort is a complicated task, which demands advance control methods. For applying such methods like model predictive control (MPC), it is necessary to have a mathematical model for the energy and lighting system in the building. This includes the model for the thermal dynamics of building besides the mathematical representations of visual comfort (DGP and workplace illuminance) as a functions of state, input and disturbance variables of system. For obtaining these models, we need informative measurement data collected by performing specific suitable experiments. These experiments should be carefully designed and implemented. Once the data is available, one can use system identification and machine learning techniques in order to obtain suitable mathematical description of the system and other essential models. For implementation of MPC, we need also the prediction of climate variables and possibly the prices of electrical power and thermal energies. Toward the formulation of control problem, it is also necessary to define a cost function that reflects the energy consumption, the monetary cost, and/or the discomfort of users.
Various issues can affect the performance of the implemented MPC such as the sampling time and horizon of MPC. However, the most important factor is the quality of derived mathematical models, i.e., if the models are not reasonably precise in predicting, the performance of MPC will not be satisfactory. Therefore, a main task in this approach is obtaining high quality models, verifying them in simulation and practice, and finally optimizing the general performance by tuning parameters.
In this project, the introduced approach will be well designed and implemented on a research unit of NEST which is a real experimental building located in Dubendorf, Zurich. This specific unit is Solace, where required measurement sensors and actuators are installed. Accomplishing the goals of this projects requires numerical and experimental steps summarized in a task list (provided in attached project description file).
In this project, the introduced approach will be well designed and implemented on a research unit of NEST which is a real experimental building located in Dubendorf, Zurich. This specific unit is Solace, where required measurement sensors and actuators are installed. Accomplishing the goals of this projects requires numerical and experimental steps summarized in a task list (provided in attached project description file).
Mohammad Khosravi -
Automatic Control Lab (IfA), ETH Zurich,
khosravm@control.ee.ethz.ch
Ahmed Aboudonia -
Automatic Control Lab (IfA), ETH Zurich,
ahmedab@control.ee.ethz.ch
Mohammad Khosravi - Automatic Control Lab (IfA), ETH Zurich,
khosravm@control.ee.ethz.ch
Ahmed Aboudonia - Automatic Control Lab (IfA), ETH Zurich,