Category: A1; Team name: SPAICOM_CattDiN; Dataset: LargeScaleMultipurposeBenchmarkDatasetsWDN

[LeoDiNino97]

Checklist

• [X] My pull request has a clear and explanatory title.
• [X] My pull request passes the Linting test.
• [X] I added appropriate unit tests and I made sure the code passes all unit tests. (refer to comment below)
• [X] My PR follows PEP8 guidelines. (refer to comment below)
• [X] My code is properly documented, using numpy docs conventions, and I made sure the documentation renders properly.
• [X] I linked to issues and PRs that are relevant to this PR.

Description

This pull requests integrates eleven datasets for Water Distribution Network (WDN) analysis as described in [1]: they are generated synthetically via numerical simulation given well-known configurations. The eleven datasets are Anytown,Balerman,C-Town,D-Town,EXN,KY1,KY6,KY8,KY13,L-Town,Modena. Each of these datasets comprise many different .csv files; however, we restricted the interest to the following files containing time-series generated as described in the reference paper:

WDN	Domain
pressure.csv	Nodes
demand.csv	Nodes
flowrate.csv	Edges
velocity.csv	Edges
head.csv	Nodes
head_loss.csv	Edges
friction_factor.csv	Edges
attrs.json	-

Each of these files includes a certain number of scenarios, each of which has a certain temporal resolution in terms of subsquent snapshots: the number of scenarios is stored under the key gen_batch_size in the attrs.json files, while the number of time-stamps is stored under the key duration. The metadata file attrs.json also contains the graph in terms of adjacency list under the key adj_list.

Water Distribution Networks (WDNs) can be naturally represented as graphs, which has led to extensive use of graph deep learning for a variety of challenging tasks [2]. These applications span both transductive settings-such as estimating the full network state from partial observations at a single time snapshot-and inductive settings, where models leverage spatiotemporal structure to perform tasks like demand forecasting. Clearly, the datasets poses many possible regression problems that can be cast at node-level, edge-level, and at a combined level.

However, the physical parameters of WDNs are governed by a variety of structural laws that impose higher-order topological constraints on the system [3]. Thus, we believe that topological deep learning could offer a powerful and principled framework for addressing the engineering problems posed by WDN monitoring and analysis.

Issue

We know that spatio-temporal and cross-domain learning is beyond the scope of the current implemented architectures of TopoBench; this in our opinion does not make such a contribution any less relevant, but rather makes it crucial in view of the implementation of new topological models for time series and for physical-informed topological learning. Moreover, it serves as the first bridge between the topological deep learning and water distribution network (WDN) communities, establishing TopoBench as a practical tool for this type of analysis.

References

[1] Tello A., et al., "Large-Scale Multipurpose Benchmark Datasets For Assessing Data-Driven Deep Learning Approaches For Water Distribution Networks" (2023)

[2] Vittori, G., et al. “Graph neural networks to model and optimize the operation of Water Distribution Networks: A review.” (2025)

[3] Cattai, T, et al. “Physics-Informed Topological Signal Processing for Water Distribution Network Monitoring” (2025)

Co-authored by @TizianaCattai and @LeoDiNino97

[levtelyatnikov]

Dear Participants,

This is a final reminder regarding the upcoming challenge deadline.

📅 Deadline: Tomorrow, 25th November 2025

✅ Critical Requirement: Please ensure your branch is passing all CI/CD tests.

If you have any pending changes, please push them and verify your build status as soon as possible.

Good luck!