Assessing the performances and transferability of graph neural network metamodels for water distribution systems
Metamodels accurately reproduce the output of physics-based hydraulic models with a significant reduction in simulation times. They are widely employed in water distribution system (WDS) analysis since they enable computationally expensive applications in the design, control, and optimisation of wat...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IWA Publishing
2023-11-01
|
| Series: | Journal of Hydroinformatics |
| Subjects: | |
| Online Access: | http://jhydro.iwaponline.com/content/25/6/2223 |
| _version_ | 1827615118687469568 |
|---|---|
| author | Bulat Kerimov Roberto Bentivoglio Alexander Garzón Elvin Isufi Franz Tscheikner-Gratl David Bernhard Steffelbauer Riccardo Taormina |
| author_facet | Bulat Kerimov Roberto Bentivoglio Alexander Garzón Elvin Isufi Franz Tscheikner-Gratl David Bernhard Steffelbauer Riccardo Taormina |
| author_sort | Bulat Kerimov |
| collection | DOAJ |
| description | Metamodels accurately reproduce the output of physics-based hydraulic models with a significant reduction in simulation times. They are widely employed in water distribution system (WDS) analysis since they enable computationally expensive applications in the design, control, and optimisation of water networks. Recent machine-learning-based metamodels grant improved fidelity and speed; however, they are only applicable to the water network they were trained on. To address this issue, we investigate graph neural networks (GNNs) as metamodels for WDSs. GNNs leverage the networked structure of WDS by learning shared coefficients and thus offering the potential of transferability. This work evaluates the suitability of GNNs as metamodels for estimating nodal pressures in steady-state EPANET simulations. We first compare the effectiveness of GNN metamodels against multi-layer perceptrons (MLPs) on several benchmark WDSs. Then, we explore the transferability of GNNs by training them concurrently on multiple WDSs. For each configuration, we calculate model accuracy and speedups with respect to the original numerical model. GNNs perform similarly to MLPs in terms of accuracy and take longer to execute but may still provide substantial speedup. Our preliminary results indicate that GNNs can learn shared representations across networks, although assessing the feasibility of truly general metamodels requires further work.
HIGHLIGHTS
The accuracy of GNN-based and MLP-based metamodels is comparable on most of the studied water networks.;
The proposed model can be trained on several water networks at once and can learn shared representation between them.;
By learning shared representations, the model achieves comparable performance while requiring fewer training examples.;
GNNs show promising results from transferability, although further study is required.; |
| first_indexed | 2024-03-09T09:06:33Z |
| format | Article |
| id | doaj.art-e6dc167defd54b20ba72810db82e213f |
| institution | Directory Open Access Journal |
| issn | 1464-7141 1465-1734 |
| language | English |
| last_indexed | 2024-03-09T09:06:33Z |
| publishDate | 2023-11-01 |
| publisher | IWA Publishing |
| record_format | Article |
| series | Journal of Hydroinformatics |
| spelling | doaj.art-e6dc167defd54b20ba72810db82e213f2023-12-02T10:27:54ZengIWA PublishingJournal of Hydroinformatics1464-71411465-17342023-11-012562223223410.2166/hydro.2023.031031Assessing the performances and transferability of graph neural network metamodels for water distribution systemsBulat Kerimov0Roberto Bentivoglio1Alexander Garzón2Elvin Isufi3Franz Tscheikner-Gratl4David Bernhard Steffelbauer5Riccardo Taormina6 Department of Civil and Environmental Engineering, Norwegian University of Science and Technology, Trondheim, Norway Department of Water Management, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands Department of Water Management, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands Department of Intelligent Systems, Faculty of Electrical Engineering, Mathematics, and Computer Science, Delft University of Technology, Delft, The Netherlands Department of Civil and Environmental Engineering, Norwegian University of Science and Technology, Trondheim, Norway Hydroinformatics Group, KWB – Kompetenzzentrum Wasser, Berlin, Germany Department of Water Management, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands Metamodels accurately reproduce the output of physics-based hydraulic models with a significant reduction in simulation times. They are widely employed in water distribution system (WDS) analysis since they enable computationally expensive applications in the design, control, and optimisation of water networks. Recent machine-learning-based metamodels grant improved fidelity and speed; however, they are only applicable to the water network they were trained on. To address this issue, we investigate graph neural networks (GNNs) as metamodels for WDSs. GNNs leverage the networked structure of WDS by learning shared coefficients and thus offering the potential of transferability. This work evaluates the suitability of GNNs as metamodels for estimating nodal pressures in steady-state EPANET simulations. We first compare the effectiveness of GNN metamodels against multi-layer perceptrons (MLPs) on several benchmark WDSs. Then, we explore the transferability of GNNs by training them concurrently on multiple WDSs. For each configuration, we calculate model accuracy and speedups with respect to the original numerical model. GNNs perform similarly to MLPs in terms of accuracy and take longer to execute but may still provide substantial speedup. Our preliminary results indicate that GNNs can learn shared representations across networks, although assessing the feasibility of truly general metamodels requires further work. HIGHLIGHTS The accuracy of GNN-based and MLP-based metamodels is comparable on most of the studied water networks.; The proposed model can be trained on several water networks at once and can learn shared representation between them.; By learning shared representations, the model achieves comparable performance while requiring fewer training examples.; GNNs show promising results from transferability, although further study is required.;http://jhydro.iwaponline.com/content/25/6/2223artificial intelligencegraph neural networksurrogate modeltransfer learningwater distribution systemwater network |
| spellingShingle | Bulat Kerimov Roberto Bentivoglio Alexander Garzón Elvin Isufi Franz Tscheikner-Gratl David Bernhard Steffelbauer Riccardo Taormina Assessing the performances and transferability of graph neural network metamodels for water distribution systems Journal of Hydroinformatics artificial intelligence graph neural network surrogate model transfer learning water distribution system water network |
| title | Assessing the performances and transferability of graph neural network metamodels for water distribution systems |
| title_full | Assessing the performances and transferability of graph neural network metamodels for water distribution systems |
| title_fullStr | Assessing the performances and transferability of graph neural network metamodels for water distribution systems |
| title_full_unstemmed | Assessing the performances and transferability of graph neural network metamodels for water distribution systems |
| title_short | Assessing the performances and transferability of graph neural network metamodels for water distribution systems |
| title_sort | assessing the performances and transferability of graph neural network metamodels for water distribution systems |
| topic | artificial intelligence graph neural network surrogate model transfer learning water distribution system water network |
| url | http://jhydro.iwaponline.com/content/25/6/2223 |
| work_keys_str_mv | AT bulatkerimov assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems AT robertobentivoglio assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems AT alexandergarzon assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems AT elvinisufi assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems AT franztscheiknergratl assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems AT davidbernhardsteffelbauer assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems AT riccardotaormina assessingtheperformancesandtransferabilityofgraphneuralnetworkmetamodelsforwaterdistributionsystems |