Congratulations to our colleague Alberto Silvestri!
We are proud to announce the successful doctorate defense of our colleague Alberto Silvestri! His PhD thesis, titled “Reinforcement learning control for energy efficient buildings” focuses on Deep Reinforcement Learning and how it can be employed to enhance the energy efficiency in buildings, in regards to Heating, Ventilation, and Air Conditioning systems.
HVAC systems ensure occupant comfort, yet they represent one of the main contributors to the overall energy consumption during building operations. Traditional control methods, such as Rule-Based Controllers, often fail to optimize building operations efficiently. Additionally, other advanced control strategies like Model Predictive Control can be challenging to implement due to the need for accurate dynamic models and substantial online computational resources. Alberto’s research conducts a comparative analysis of the performance of various DRL controllers against traditional RBC and MPC. By deploying a DRL controller in a real office building at the HiLo unit at NEST, his thesis validates the simulation results and highlights the practical challenges encountered during the implementation.
During his PhD, Alberto published the following papers that you can find and read on the ETH Zurich Research Collection:
- Silvestri, A., Lydon, G. P., Waibel, C., Wu, D. & Schlueter, A. Data-driven reduced order modelling using clusters of thermal dynamics. in Proceedings of Building Simulation 2023: 18th Conference of IBPSA 18 (IBPSA, Bruges, Belgium, 2023), 1145. (Link).
- Silvestri, A., Coraci, D., Wu, D., Borkowski, E. & Schlueter, A. Comparison of two deep reinforcement learning algorithms towards an optimal policy for smart building thermal control. Journal of Physics: Conference Series 2600, 072011 (2023). (Link).
Further papers are under review at the moment.
Thank you to the PhD supervisors Prof. Dr. Arno Schlueter, and the external examiners Prof. Dr. Massimo Fiorentini, Aarhus University, Prof. Dr. Melanie Zeilinger, Department of Mechanical and Process Engineering (D-MAVT), ETH Zurich and Prof. Dr. Prof. Dr. Silke Langenberg, ETH Zürich Institute of Technology in Architecture, ETH Zurich.
We would also like to thank Mitsubishi Electric R&D Centre Europe (MERCE-UK) for supporting and funding this research.
We wish Alberto all the best for his future!