Research and development
Fluvius
A software application based on Machine Learning algorithms designed to forecast water flow in river basins.
The problem
Climate change is making traditional water resource management systems increasingly unreliable. The erratic climate of recent years — ever longer droughts alternating with intense rainfall events — is putting a strain on water and hydropower infrastructure, redefining how water must be managed and used.
In mountain areas, characterised by large elevation changes and complex hydrology, knowing river flows in advance is crucial information from two distinct perspectives: on one hand, it allows hydropower producers to precisely plan production and reservoir management; on the other, it enables timely action to protect the area in the event of sudden floods, reducing hydrogeological risk.
The tools available today, however, struggle to keep up. Traditional physical models are expensive, complex to build and often out of reach for small and medium-sized producers; the most common statistical models, meanwhile, are built individually for each basin, without accounting for the hydrological — spatial and topological — relationships that connect neighbouring rivers.
How it works
Fluvius was created to overcome these limitations with a different approach: rather than building an isolated model for each watercourse, it represents an entire area’s river network as a single interconnected system, in which each monitored river is linked to the others based on its location and hydrological relationships.
This structure allows the system to learn not only the behaviour of each individual basin, but also the correlations between neighbouring basins, providing more accurate and consistent forecasts on a regional scale, rather than river by river.
The model combines historical flow data with weather forecasts, corrected in real time based on physical measurements from ground stations, and updates estimates hourly over a 48-hour forecast horizon — a level of resolution that makes it a genuinely operational diagnostic tool.
The system is designed to grow over time: new basins and new monitoring stations can be added without having to rebuild the model from scratch.
The impact
For hydropower producers, accurately forecasting a basin’s water availability means being able to plan production further in advance, optimise reservoir management and act with better timing on the energy market — a direct advantage especially for small and medium-sized plants, which today lack access to advanced forecasting tools.
For the protection of the area, knowing flow trends in advance means being able to act before a flood becomes an emergency.
Fluvius stems from the scientific research carried out internally by the AIAQUA team, a spin-off of the Free University of Bozen-Bolzano, with the aim of turning techniques still uncommon in the market into a concrete, accessible tool.
Project sheet
Want to know more about Fluvius?