Fraunhofer IEE reduces forecast errors for solar eclipse

The integration of photovoltaics into power grids requires increasingly better forecasts, especially with increasing capacities, because uninterrupted grid operation does not accept avoidable errors. In this context, rare meteorological events still present particular challenges. In addition to volcanic ash and Saharan dust, a solar eclipse requires the attention of grid operators, electricity traders and forecasting service providers. At the Fraunhofer Institute for Energy Economics and Energy System Technology IEE, a solution has now been developed to combine the degree of occultation with established weather forecast models specific to the location and time.

The next partial solar eclipse in Central Europe will be observable on October 25th, 2022, and will significantly reduce photovoltaic (PV) feed-in from solar plants and solar parks. So far, weather forecasts do not routinely include such rare extreme events. Therefore, experts at Fraunhofer IEE have developed and validated a flexible solution to be able to combine the degree of coverage specifically for location and time with all important weather forecasts. In the future, this will allow regional and local PV feed-in forecasts to be optimally adjusted and errors to be reduced.

The previous estimation of a solar eclipse must be possible entirely based on previously known data. Already through persistence forecast (clear-sky index) the diurnal variation of global radiation or PV feed-in can be continued well over a solar eclipse. Although the effect of the solar eclipse can be slower and smaller than the influence of changing cloud cover, it plays a significant role in the results. Volatile self-consumption also has a massive effect on the feed-in profile of solar plants and therefore can and must be treated separately

The validation of the results with data from June 10, 2021, was conducted in two steps: by ground-based measurement of global radiation based on data from the German Weather Service DWD and by feed-in measurements of thousands of PV parks, which were used for forecasting and estimation processes at German TSO Amprion GmbH.

Weather and power forecasts from Fraunhofer IEE already include the new simulation feature. In addition, the solution is directly available as an algorithm or ready-to-use software module for user integration. This avoids errors and optimally supports marketing and grid operation.

Optimized power forecasts at new locations

Power forecasts are a fundamental element for secure grid integration of renewable energies. The further expansion of PV plants in Germany requires well-scalable models that can, shortly after plant installation, provide forecasts of high quality. Using multi-task- and transfer-learning, this requirement has been implemented in a methodology and successfully tested at Fraunhofer IEE for PV forecasts as well as (self-) consumption forecasts.

Alternatively, extrapolations can be made to new, unknown or unmeasured locations and regions, based on existing solar plants. Such projection methods for real-time measurement data and forecasts were optimized and compared in the SOLREV project in cooperation with Fraunhofer ISE and the German transmission system operators.

Solar eclipse in October 2022 with strong occultation

At about 25% coverage, the solar eclipse on October 25, 2022, will be about twice as strong as the solar eclipse on June 10, 2021, at about 13%. Detailed weather forecasts a few days before will provide information on how strong the cloud cover and thus ultimately the effect of the solar eclipse will be.

Overall, the study results are promising and allow the application to individual weather forecasts. "In addition to the already strong forecasting models for the feed-in of individual PV parks or entire portfolios - including self-consumption and other individual effects - this can make an important contribution to avoiding or reducing errors," says Dr. Rafael Fritz of Fraunhofer IEE. Data from each following solar eclipse and other optimizations will further improve the methodology and make energy systems worldwide more resilient to such locally, but not globally, rare extreme events.

The results of the simulation of solar eclipses for use in PV feed-in forecasts will be part of an Expert Web Session at Fraunhofer IEE in October.

Research contact:
Dr. Rafael Fritz, Fraunhofer IEE, Product Manager Solar Power Forecasts

Last modified: