HYFLEX
AI-powered hybrid dispatch coordination platform for small hydropower and hybrid energy sites
Europe has more than 20,000 small hydropower plants, many of which are adding PV, batteries, heat pumps and EV charging to stabilise revenues and provide flexibility to the grid. However, sub-10 MW operators lack an affordable platform that can coordinate these assets together: utility-scale DERMS are too expensive and complex, while small-hydro controllers focus on turbine operation and do not optimise hybrid sites.
As a result, hybridised small hydropower sites are often managed through separate device-level rules, leading to water spillage, uncoordinated peaks and missed flexibility value. The technical barrier is the real-time stochastic dispatch problem: operators must coordinate uncertain river inflows, PV generation, battery state-of-charge, flexible loads, ecological constraints and grid signals within short control cycles and on legacy SCADA infrastructure.
HYFLEX will deliver a stochastic model-predictive control platform that coordinates hydro units, batteries, PV and flexible loads in one optimisation loop. It will use HBV-class hydrological forecasting with numerical weather prediction ensembles, scenario-based mixed-integer optimisation, rolling-horizon re-optimisation and rule-based fallback to operate within practical control-cycle limits at small hydropower sites.
An AI closed-loop learning module will continuously improve hydrological and generation forecasts from operational feedback. During the project it will learn from the Mošenik pilot data; after commercial deployment, anonymised patterns from multiple licensed sites can strengthen shared predictive models and improve dispatch quality across the fleet.
The system will integrate via a protocol-agnostic API supporting Modbus, IEC 60870-5-104 and OPC-UA, enabling deployment as an overlay rather than a control-system replacement. It will also include ecological co-optimisation, scheduling decisions such as sediment flushing and fish passage windows while generating auditable evidence for concession and compliance processes.
Nites will lead development of the HYFLEX platform, including the stochastic MPC optimiser, hydrological forecasting module, multi-asset coordinator, protocol-agnostic SCADA/EMS integration layer and open API specification. Pareto AI will develop the active learning pipeline that retrains forecast and generation models from operational feedback and establishes the architecture for future cross-site learning from anonymised deployment data.
BPT will host the operational pilot at the Mošenik cascade in Slovenia, providing SCADA access, operational data, hydrological ground truth and operator feedback. The consortium will validate live hydro dispatch over at least six months against rule-based operation, targeting at least 10% higher energy yield per m³ and 20% lower spillage, while using a calibrated digital twin to test PV, BESS, heat pump and EV charging coordination with a target of at least 5% revenue uplift.
The project will also implement ecological co-optimisation and automated audit evidence, publish a versioned open API specification, create a public performance dashboard for the pilot site, assess replication across different site typologies, and prepare commercial roll-out through early-adopter letters of intent and integrator-ready deployment documentation.
.png)
