A Toolbox for Supporting Research on AI in Water Distribution Networks
André Artelt, Marios S. Kyriakou, Stelios G. Vrachimis, Demetrios G. Eliades, Barbara Hammer, Marios M. Polycarpou
TL;DR
This paper addresses the challenge of applying AI in Water Distribution Networks under uncertainty by introducing EPyT-Flow, a Python toolbox designed for realistic scenario generation and streamlined access to benchmark datasets. Built on EPANET/EPANET-MSX, EPyT-Flow provides 16 ready-to-use networks, 7 benchmarks with evaluation metrics, a residual-based sensor interpolation detector as a baseline, and a gym-like environment for reinforcement learning-based control, along with 4 event types and 11 uncertainty types. The authors demonstrate the toolkit with a use-case on the L-Town network, showing that the baseline detector can identify abrupt leaks and sensor faults but struggles with incipient leaks, highlighting both utility and areas for improvement. Collectively, EPyT-Flow supports reproducible AI research in WDNs and aims to catalyze the development of data generation, benchmarks, and AI models through BenchmarkHub and ModelHub components.
Abstract
Drinking water is a vital resource for humanity, and thus, Water Distribution Networks (WDNs) are considered critical infrastructures in modern societies. The operation of WDNs is subject to diverse challenges such as water leakages and contamination, cyber/physical attacks, high energy consumption during pump operation, etc. With model-based methods reaching their limits due to various uncertainty sources, AI methods offer promising solutions to those challenges. In this work, we introduce a Python toolbox for complex scenario modeling \& generation such that AI researchers can easily access challenging problems from the drinking water domain. Besides providing a high-level interface for the easy generation of hydraulic and water quality scenario data, it also provides easy access to popular event detection benchmarks and an environment for developing control algorithms.