An Overview of MAGNets: A Python package for Model reduction and AGgregation of water distribution Networks

Hydraulic models of water distribution systems can contain thousands of elements. Running simulations of large hydraulic models, especially to test different operational scenarios on demand, can consume excessive computational time. Therefore, reducing the size of the models by removing unessential components, while simultaneously maintaining the fidelity of the system’s hydraulic performance, enables users to run simulations quickly and correctly for large network models. Model reduction has been widely adopted for a range of water distribution modeling purposes, and yet few tools are available to reduce models on command. This work highlights MAGNets, an open-source Python package that is capable of aggregating and reducing EPANET-compatible water network models. The reduction algorithm is based on the method of variable elimination previously proposed in the literature, and involves the: (i) linearization of the system hydraulic equations at a specific operating point, (ii) iterative aggregation of nodal demands, update of pipe characteristics, and elimination of unessential nodes, and (iii) conversion of the reduced linearized model back to its original non-linear form. MAGNets allows users to customize the reduction of a network model by choosing the operating point, specifying which nodes must remain in the model, and specifying the maximum nodal degree of nodes removed. Results from testing different node removal strategies imply that a dynamic ordering of nodes outperforms static and random orderings in terms of run times and model complexity. We use the CTown network to demonstrate the application of MAGNets to analyze a system with multiple demand patterns. The Python package includes 12 benchmark networks for testing and validation, as well as examples illustrating different functionalities. MAGNets enables the efficient reduction of hydraulic models in order to perform critical tasks, such as state estimation and control scheduling on demand.