DiTEC-WDN Dataset (DWD)

DWD is a collection of synthetic, simulated and steady-state scenarios derived from 36 Water Distribution Networks (WDNs).

Each network has 1,000 distinct hydraulic scenarios generated by an open-sourced simulation toolkit EPANET.

We then describe surronding concepts as follows:

  • Scenario denotes as a sequence of snapshots.

  • Snapshot represents as a network's state often modelled as a graph at a particular timestep.

  • Nodes models a reservoir, junction, or tank in the snapshot graph. Each type has same properties and unique ones.

  • Edges refers pipe, pump, or valve in the snapshot graph. Each type has same properties and unique ones.

  • Input parameters involves in simulation input parameters, such as demands, pipe diameter, and so on.

  • Output parameters includes simulation measurements (e.g., pressure, flow rate, head, ...)

Both parameters are described as nodal/edge features in the snapshot graph. Their values are diverse but temporally correlated with those of other snapshots in the same scenario. However, in DWD, any two scenarios are independent and unrelated since they are created from different configurations (despite the same original network).

DWD is designed to:

  • Promote open scientific research in the water domain.

  • Eliminate risks of exposing sensitive data, privacy issues, or safety concerns.

  • Provide a benchmark for data-driven machine learning methods and large-scale scenario analysis.

This wiki details the process of creating DWD, including parameter optimization, simulation, and encapsulation.

It also explains how to use DWD’s data interface, GiDA.

With just two inputs, an .INP file and a YAML configuration, you can generate diverse scenarios and apply this to your own private WDN.

Acknowledgement

This work is funded by the project DiTEC: Digital Twin for Evolutionary Changes in Water Networks (NWO 19454).

Citing DiTEC-WDN

If you use the dataset, please cite:


@misc{huy2025dwd}{
    title={DiTEC-WDN: A Large-Scale Dataset of Water Distribution Network Scenarios under Diverse Hydraulic Conditions}, 
    author={Huy Truong and Andr\'{e}s Tello and Alexander Lazovik and Victoria Degeler},
    year={2025},
    note = {HT and AT contributed equally to this work. The dataset is linked to a paper submitted to *Nature Scientific Data*.}
}