Probabilistic photovoltaic power forecasts across Switzerland, powered by 1 km numerical weather prediction and bottom-up PV modelling, updated throughout the day for intraday and day-ahead energy decisions.
As photovoltaic capacity continues to expand, increasingly variable solar generation creates new challenges for energy markets and grid stability. Our mission is to provide high-precision, probabilistic photovoltaic power forecasts that help market participants anticipate generation, manage uncertainty, and make better-informed trading and electricity pricing decisions.
Latest Swiss PV generation forecast, updated with every new numerical weather prediction release.
Our forecasting system combines high-resolution numerical weather prediction, bottom-up physical PV modelling and machine-learning post-processing to translate atmospheric uncertainty into production uncertainty.
We use high-resolution 1 km ICON ensemble forecasts provided by MeteoSwiss to capture the spatial and temporal evolution of atmospheric conditions across Switzerland.
We simulate PV production from the individual installations contained in the Swiss panel registry rather than treating Switzerland as a single aggregated system.
Estimated panel orientation and tilt are retrieved from the Swiss Federal Office of Energy's Suitability of roofs for the use of solar energy dataset, allowing us to model the effects of changing solar angles and direct irradiance on PV production. An in-house snow correction further accounts for production losses associated with snow cover.
The panel dataset is regularly updated as new installations are added, allowing the forecast to evolve with Switzerland's growing PV fleet and changing generation profile.
Each ICON ensemble member is independently passed through the panel-level PV simulation. The resulting production forecasts are aggregated across the Swiss PV fleet and cantons.
This directly translates uncertainty in the weather forecast into a range of possible PV production outcomes.
Machine-learning post-processing learns systematic relationships between modelled and observed production that are difficult to represent explicitly in the physical model.
This helps account for factors such as soiling, changing panel conditions and other unmodelled system effects, while adapting to shifting systematic forecast bias over time.
The forecasting system evolves with every new weather forecast and every new observation of Swiss PV production.
Forecasts are refreshed following the ICON release schedule, keeping the latest available numerical weather prediction information in the system throughout the day.
When new Pronovo gross PV production data become available, they are incorporated into the machine-learning post-processing to continuously adapt to changing forecast biases and system conditions.
Forecast errors are evaluated against observed Pronovo production data and published when new observations become available, allowing users to understand model performance and forecast uncertainty when making decisions.
Probabilistic solar forecasts provide actionable insight into generation, uncertainty and regional production for energy market participants and grid operators.
Improve expectations of upcoming solar generation and forecast uncertainty to support more informed bidding strategies. Probabilistic forecasts enable low- and high-generation scenarios to be incorporated according to a company's energy reserves and risk tolerance across day-ahead and intraday spot markets, as well as balancing and control energy markets.
Our bottom-up approach allows PV generation to be aggregated for specific plant portfolios or geographical regions, providing forecasts tailored to the assets being managed.
Virtual power plants and flexibility pools containing solar assets can use these forecasts to anticipate available generation and participate more effectively in energy markets. The P10 forecast provides an additional low-generation scenario for risk-aware bidding and portfolio management.
Forecast accuracy can be further improved when historical generation data are available. Customers can also provide their own installed capacity, geographical coordinates, tilt and orientation, allowing their assets to be directly incorporated into the forecasting pipeline.
Cantonal and national PV production forecasts provide grid operators with visibility into expected generation levels, uncertainty and potential fluctuations, supporting grid stability and operational planning.
Because the physical model is built from individual installations, PV generation can also be aggregated geographically to provide regional or substation-level forecasts, offering more granular visibility into where solar generation may affect the grid.
Our forecasting system combines Swiss meteorological, geospatial and PV production datasets with European electricity-market data.
High-resolution numerical weather forecasts from the MeteoSwiss ICON-CH1/2-EPS ensemble provide the atmospheric inputs for the PV simulations.
The Swiss Federal Office of Energy registry provides the installation-level basis for our bottom-up representation of the Swiss PV fleet. The dataset is regularly updated so that new installations can be incorporated as the national PV fleet grows.
Roof suitability data are used to estimate the orientation and tilt characteristics of PV installations, improving the physical representation of the installed PV fleet.
Pronovo provides aggregated cantonal gross PV production profiles from installations with available load profiles. These observations form the basis for constructing an estimated total cantonal PV production reference.
European electricity-system data are used for comparison and benchmarking against published generation forecasts and observed generation for wind and solar.
The data providers referenced on this page are independent third-party sources and are not affiliated with, do not endorse, and are not responsible for this project, its analysis, methodology or conclusions.
In particular, the use of Pronovo production data and the scaling assumptions applied to estimate total PV production do not represent an endorsement, validation or responsibility by Pronovo for the resulting analysis or forecasts.
Pronovo provides aggregated cantonal gross production profiles from PV installations with available load profiles. However, not all registered installations provide a load profile.
To represent the full Swiss PV fleet, the available Pronovo production profiles are scaled for each canton according to the total registered PV capacity and the capacity represented in the corresponding Pronovo gross production profile.
The resulting series provides an estimated total cantonal PV production reference that accounts for installations not represented in the available load-profile data. This estimated reference is used for machine-learning calibration and out-of-sample forecast evaluation.
The current system focuses on reliable next-day forecasting, with a longer horizon being developed.
Swiss canton-level forecasts generated from the current ensemble forecasting system.
Extension using a deterministic local forecast to provide a longer operational forecasting horizon.
View the latest canton-level production forecasts, forecast uncertainty and regional solar conditions.
Open Dashboard →