Wit, Bram de (2021) From Data To Control: Learning An Optimal HVAC Control Policy Using A Deep Learning Based Neural Twin. Master's Thesis / Essay, Artificial Intelligence.
|
Text
mAI_2021_deWitB.pdf Download (6MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (120kB) |
Abstract
The building and construction sector is responsible for 36% of the total energy usage and for 39% of energy/process related emissions. A large part of this energy is used by heating, ventilating and air conditioning (HVAC) systems that regulate a building's indoor climate. This thesis aims at developing smart HVAC controllers that can be used at scale. The proposed controllers achieve this by deriving the control solely from data. First a Neural Twin, which serves as a simulator of the process, is trained. The results show that the Neural Twin framework is able to simulate several distinct processes with an average absolute error close to 0.2°C for all processes. The Neural Twin is then used to develop two different control algorithms. One control algorithm uses Proximal Policy Optimization (PPO) and the other control algorithm uses Model-Predictive Control (MPC). The results show that the control algorithms are able to handle a wide variety of processes, without manual tuning. The controller trained using reinforcement learning showed the best performance. It is estimated that the proposed controller can lead to a 5% - 40% decrease in effective energy usage, while retaining the thermal comfort and stability. From the results it is concluded that the control methods pose an attractive alternative compared to the conventional controllers.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Schomaker, L.R.B. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 15 Nov 2021 13:27 |
| Last Modified: | 15 Nov 2021 13:27 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/26284 |
Actions (login required)
 |
View Item |