District Heating Networks (DHNs) are flexibile and efficient heat delivery systems. Through the last decades, these systems have gained popular interests from European contries. DHNs provide low cost solutions to incorporate renewable sources in heat production and contributing to CO2 reductions inline with the EU 2050 climat-neutral objective.
Major problematic faced by DHNs operators come from the high computational costs of modeling these systems. The main reason is that these networks are spatially sparse. Therefore, a reduction of such sparsity is the straightforward solution to this problem.
In our paper, we propose an unsupervised clustering-based approach to identify `similar' substation nodes which are replaced by trained AI models. A task-driven distance metric has been introduced to quantify such similarty.
This repository contains the obtained results presented in our paper, the data used and the codes for reproducibility.
The study investigates the performance of the proposed approach on 17 different DHNs (i.e., sizes, thermal demands, etc.). The topology files and visualisation figures can be found in the studied DHNs folder. In our approach, the DHN is viewed as one-layered directed graph with times series signals on the nodes and edges.
- Nodes represent heat consumers, heat producers or pipes junctions points
- Edges represent both identifical supply and return pipes oriented along the supply pipes
Edges are colored in black. However, we indicate in red dashed the edges which pipes have flux inversions.
To install all required packages, enter the following line commands at the project source folder
python -m pip install -r requirements.txtAll code sources can be found in the source folder. An example of performing the clustering is proposed in the proposed example.ipynb which presents in great details how to perform the clustering. We recall to the readers that the physical simulation states data of the DHNs are not available in this repository due to their memory sizes. However, they can be made available at request.
We provide an illustration example in the notebook example. This illustration uses the DHN with indicator 1.
We welcome all comments and improvements contributions. Please contact us or submit an issue.
Data and codes in this project are protected under the European Union Public Licence (EUPL) v1.2. For more information see LICENSE.