For Primeo Energie, a Swiss energy supplier that manages around one gigawatt of installed PV capacity, solar radiation is by far the most important weather parameter. The company handles the marketing and short-term optimization of more than ten thousand plants, most of them small rooftop systems, spread across the country. Each year, balancing energy runs to tens of millions of Swiss francs.

In the past, Primeo's forecasting looked much like everyone else's in the industry. The team pulled a radiation forecast about once a day for a handful of sites, using legacy tools, and corrected it by hand whenever the weather shifted. Given the volume of PV Primeo manages, that approach had reached its limits.

"Since Switzerland moved to a single-price model for balancing energy, costs have come down somewhat, but on days with highly variable weather things can still get expensive very quickly. A single weekend can account for half of the month's balancing energy costs, and that's exactly when we have fewer people working. That was one of the main reasons we decided this had to run automatically," explains Tobias Ott, Head of BGM and Analytics at Primeo.

Today, Primeo pulls data for around 150 representative locations in Switzerland at every weather model update and recalculates all forecasts automatically, every three hours for ICON-D2 and every hour for EURO1k. Within ten minutes, a fresh PV forecast is ready for the intraday trading desk.

Update frequency

For intraday trading, what Primeo's traders value most is frequency. EURO1k updates every hour, whereas many other models like ICON-D2 update only every three hours, often at awkward moments. On a tricky day, the picture at 8 a.m. is still unclear, and an update arriving around 11:30 a.m. can change the whole afternoon, by which point it's often too late to act on it. EURO1k's hourly updates close that gap.

Independent view

EURO1k is also a genuinely different model, not built on the same ICON core as Primeo's other inputs. That gives traders an independent second opinion and reduces the risk of every forecast being wrong in the same direction at once. As Tobias puts it, "Very similar models tend to be wrong in the same direction. Having several independent views from different weather models helps our traders make far more robust decisions."

Complete coverage of Europe

EURO1k also covers all of Europe with a consistent dataset, which has become a strategic factor for Primeo. For a company that is already active in Switzerland and France and planning to expand into other countries, that consistency is a real advantage over geographically limited models. With EURO1k, the same model can be used everywhere Primeo operates, and the team can extend into other countries without evaluating providers from scratch or stitching together a patchwork of regional sources.

Data delivery reliability

With forecasts running automatically around the clock, the setup is only as good as the API that feeds it. Primeo depends on data arriving reliably and fast, in large volumes, every day of the week. "Because we pull so many forecasts automatically, we depend on the data coming through the API reliably and quickly, in large volumes, on weekdays and weekends alike. Meteomatics does an excellent job there," says Tobias.

Since rebuilding its data setup around the Meteomatics Weather API, Primeo has reduced its overall PV forecast error by about 40%. "Every percent of improvement translates directly into euros, which for a portfolio like ours means millions of euros a year," says Tobias.

Besides reduction of imbalance cost, accurate forecasts of PV assets are also important for marketing of these sites on the balancing market.

Switzerland's PV fleet will keep growing, and more of those assets will be curtailable at short notice. Managing them depends on accurate forecasts, both day-ahead and intraday, and Primeo's automated setup is built to absorb that growth: adding sites or pulling new parameters is a matter of configuration through the API, not new hardware or IT projects.

"We've taken a big step forward. We now get frequent updates from a large number of sites and can process them all automatically. I think that's the foundation we need to manage so many renewable assets in the future," says Tobias.