TwinAI-demo. Real-time digital twin and graph reinforcement learning for interactive water distribution network management

This demo presents the functioning of TwinAI, an interactive framework for real-time management of Water Distribution Networks (WDNs) that integrates a dynamic digital twin with a graph-based reinforcement learning agent. Unlike traditional offline simulators, TwinAI enables live interaction with an ongoing hydraulic simulation, allowing operators to visualize network states, inject anomalies, and execute control actions directly from a Graphical User Interface (GUI). The demo showcases real-time operations, leak isolation, simulation branching for what-if analysis, and autonomous decision-making driven by a Graph-RL agent. The system demonstrates how physically consistent digital twins and learning-based control can be combined for adaptive WDN operation and for interactive what-if analyses.