Register now After registration you will be able to apply for this opportunity online.
Robot-Assisted Building Energy Management
The building sector is responsible for more than one third of the energy consumption and CO2-emissions worldwide. In industrialized countries around half of this energy is used for heating, ventilation and air conditioning (HVAC). Improving the energy efficiency of buildings has therefore a significant impact on mitigation climate change. However, renovating the building envelope or HVAC system of already existing buildings is relatively expensive and causes slow diffusion. In contrast, integration or upgrading of heating or cooling control systems and optimizing its operation can be done at comparatively low costs, resulting in fast impact.
While Model Predictive Control (MPC) is considered the gold standard for climate control in buildings, a crucial part of MPC is the identification of an accurate model of the building. Here, first-principle based models still outperform purely data-driven models such as the ones presented in. However, first-principle based building modelling is associated with many tedious tasks such as mapping of the floor plan and inventory, or identification of several system parameters (e.g. thermal resistance of walls). If an experienced engineer has to perform these tasks for each building individually, MPC might not be economically feasible.
Keywords: MPC, Optimization, Machine Learning, Sensor fusion, Robotics, Buildings, Energy
Within the Robot-Assisted Building Energy Management project, the student will develop and test a set of methods to identify dynamic building models and estimate system states which potentially can be performed by a robot platform (e.g. room geometry mapping using LiDAR sensors, estimating the thermal resistance of walls with the help of IR cameras, or measuring temperature and indoor air quality applying sensor fusion techniques). This first project phase might be an iterative process in collaboration with the robotics team. The output of this phase is expected to be a set of methods for control-oriented dynamic building models identification.
In the second phase of the project, the model identification methods will be applied in Empa’s NEST building, the accuracy of the resulting models compared to each other and against state-of-the-art data-driven building models. Moreover, the increasing associated costs (i.e. additional costs of sensors) also need to be weighted against the model accuracy improvement in order to find the best solution. Finally, Model Predictive Control (MPC) for energy efficient climate control will be implemented using the most promising model identified and tested at NEST in real-life.
Within the Robot-Assisted Building Energy Management project, the student will develop and test a set of methods to identify dynamic building models and estimate system states which potentially can be performed by a robot platform (e.g. room geometry mapping using LiDAR sensors, estimating the thermal resistance of walls with the help of IR cameras, or measuring temperature and indoor air quality applying sensor fusion techniques). This first project phase might be an iterative process in collaboration with the robotics team. The output of this phase is expected to be a set of methods for control-oriented dynamic building models identification. In the second phase of the project, the model identification methods will be applied in Empa’s NEST building, the accuracy of the resulting models compared to each other and against state-of-the-art data-driven building models. Moreover, the increasing associated costs (i.e. additional costs of sensors) also need to be weighted against the model accuracy improvement in order to find the best solution. Finally, Model Predictive Control (MPC) for energy efficient climate control will be implemented using the most promising model identified and tested at NEST in real-life.
Develop, validate, and compare a different methods to identify dynamic building models and estimate system states that can potentially be performed by a robot platform (e.g. room geometry mapping using LiDAR sensors, estimating the thermal resistance of walls with the help of IR cameras, or measuring temperature and indoor air quality applying sensor fusion).
Implement and apply Model Predictive Control (MPC) in real-life using the most promising model.
Develop, validate, and compare a different methods to identify dynamic building models and estimate system states that can potentially be performed by a robot platform (e.g. room geometry mapping using LiDAR sensors, estimating the thermal resistance of walls with the help of IR cameras, or measuring temperature and indoor air quality applying sensor fusion). Implement and apply Model Predictive Control (MPC) in real-life using the most promising model.