Safe Contextual Bayesian Optimization for Sustainable Room Temperature PID Control Tuning
Authors: Marcello Fiducioso, Sebastian Curi, Benedikt Schumacher, Markus Gwerder, Andreas Krause
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this work, we use a simulator based on Gao et al. [2007] provided by Siemens Building Technology Division.2. The model handles thermal gains due to solar radiation, people presence and equipment heating. The heat exchange with the environment varies according to the outside air temperature. The room control loop regulates the amount of radiator inlet water. The water temperature control loop uses the automatic weather compensation system, which determines the water temperature according to the outside air temperature. Our goal is to optimize the room control loop without knowledge of the water temperature nor its set-point. |
| Researcher Affiliation | Collaboration | 1 Department of Computer Science, ETH Zurich, Switzerland 2 Building Technologies Division, Siemens Switzerland Ltd. |
| Pseudocode | Yes | Algorithm 1: Safe contextual GP-LCB. |
| Open Source Code | No | The paper does not provide any link to open-source code or state that code is available in supplementary materials or upon request. |
| Open Datasets | No | The paper uses a simulator based on real measurements from Zug, Switzerland, but does not provide public access to the raw measurements or a specific dataset for training. The simulator itself is provided by Siemens Building Technology Division, but no public link is given. |
| Dataset Splits | No | The paper describes running a continuous simulation over a "heating season" to tune parameters and evaluate performance. It does not mention or provide explicit train/validation/test dataset splits as would be typical for static machine learning datasets. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as CPU/GPU models, memory, or cloud instance types used for running the experiments. It only mentions the use of a simulator. |
| Software Dependencies | No | The paper does not list any software dependencies with specific version numbers (e.g., Python 3.x, PyTorch 1.x, specific libraries or solvers with versions). |
| Experiment Setup | Yes | The algorithm exploratory parameter is fixed to β = 2 as it proved to have the best performance in our simulations. The costs are all normalized so that in 95% of the historical data, they are between 0 and 1. In this work, we select all the weights equal to 0.25 to normalize the total cost between 0 and 1. Similarly, we set the parameters ci so that 97.5% of each of the historical costs are considered safe. |