The platform is under development. Coming soon!
Impact

Better forecasts can improve both system operation and the understanding of distributed solar generation.

SentinelPV combines scientific advances with a demonstrator intended for grid operators, energy communities, storage managers, researchers and prosumers.

System and market impact

More visibility on distributed generation.

Distributed PV is difficult to operate when its current output and near-term evolution are only partially visible. By reconstructing and forecasting production with greater spatial detail, SentinelPV can support day-ahead scheduling, grid balancing and the management of uncertainty associated with variable renewable generation.

For transmission and distribution system operators, improved forecasts can provide better situational awareness of the expected contribution of PV generation and of the areas in which larger deviations may occur. For energy communities and storage operators, the same information can support the scheduling of flexible resources, charge and discharge decisions and the coordination of local generation and consumption.

The platform is not intended to replace operational control-room tools during the research project. Its role is to demonstrate how a representative monitoring network and hybrid forecasting methodology can provide information that can later be integrated into more advanced decision-support systems.

Illustration showing the link between distributed PV forecasts, grid balancing, storage and energy communities
Scientific and technological impact

Connecting physical knowledge with data-driven forecasting.

The project contributes to a research direction in which photovoltaic models and machine learning are not treated as competing alternatives, but as complementary components.

Hybrid modelling and spatial learning

Physical models provide interpretable relationships between irradiance, temperature and PV conversion, while machine learning can capture residual nonlinearities and spatio-temporal patterns. The sentinel concept adds a further research layer by linking model performance to the design of the monitoring network itself.

The project therefore generates knowledge not only on forecasting algorithms, but also on the spatial resolution and observation density required for regional PV forecasting.

Benchmarking and reproducibility

A common benchmark makes different forecasting approaches comparable under the same conditions. Performance is analysed through numerical metrics, time series and error distributions, including challenging atmospheric cases rather than only average days.

Where possible, methodological documentation, selected data products and software components can support reuse by the research community and future development beyond the project.

Stakeholders and society

Forecast information at different levels of technical detail.

The same forecasting architecture can serve different users when information is presented with the appropriate level of detail.

Grid and energy operators

Quantitative forecasts, model comparison and spatial information can support planning and analysis by TSOs, DSOs, energy service providers and storage operators.

Energy communities and prosumers

Clear geographical summaries of historical and expected PV production can help local users understand resource variability and coordinate consumption, generation and storage more effectively.

Research and future projects

The modular architecture is designed to be transferable to other regions and to form the basis for broader collaborative research on renewable forecasting, including future extensions to wind generation and cross-border applications.