Digital Twins and Edge AI for Small/Off-Grid Water Systems: A Review

Here, we review compiled field evidence on Digital Twins and Edge AI for small off-grid water systems. We examined 28 studies from 2018-2025, including 19 peer-reviewed papers and 9 gray literature sources. All included works reported experimental pilots, edge deployments, or testbeds implemented in decentralized environments. Three reference architectures were identified: edge-only DTs for ultra-off-grid settings, federated DT networks for community-scale coordination, and hybrid edge-cloud DTs for intermittent connectivity. Common hardware included Raspberry Pi Zero, sub-$15 sensors, and ESP32, typically operating within 10-30 Wh/ day energy budgets powered by solar systems. Results consistently indicated that simplified models outperform complex alternatives. Reduced-order physics models form the foundation, and lightweight ML correctors make them adaptable. Deployed models remain static and are updated seasonally using Bayesian recalibration. Popular uses include atmospheric water harvesting, containerized wastewater treatment, leak detection, brackish water filtration, and small-scale irrigation. Empirical results across pilots were consistent: energy or water consumption dropped 18-40%, downtime or repair time fell by 27-55%, and component life grew by over four months. This review proposes a standard comparison tool for future studies, which identifies key nontechnical problems and presents a practical guide for field trials. We outline a research pathway for self-healing models, including discussions on shared standards, integration of social and technical aspects, and long-term testbed development. Researchers, policymakers, and NGOs may find clear, actionable steps for improving access to reliable and affordable water services in resource-limited areas.