Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Optimisation Method for the Clear Sky PV Forecast Using Power Records from Arbitrarily Oriented Panels

: Thomas, Jorge A.


Institute of Electrical and Electronics Engineers -IEEE-; IEEE Industry Applications Society; IEEE Industrial Electronics Society -IES-:
7th International IEEE Conference on Renewable Energy Research and Applications, ICRERA 2018 : Paris, France, 14-17 October 2018
Piscataway, NJ: IEEE, 2018
ISBN: 978-1-5386-5982-3
ISBN: 978-1-5386-5981-6
ISBN: 978-1-5386-5983-0
International Conference on Renewable Energy Research and Applications (ICRERA) <7, 2018, Paris>
Fraunhofer IOSB ()
photovoltaic system; energy forecast; turbidity factor; optimisation; machine learning

Since arbitrarily oriented solar panels are everywhere, the present work develops an algorithm to estimate the on-site turbidity variation in order to optimise the clear sky photovoltaic forecast, using past power records produced in presence of the isotropic diffuse irradiance component during absence of clouds when the generator is under full load; these data are considered as training datasets. The algorithm is based on Ineichen-Perez’s solar resource and separation model, where turbidity factors are treated as optimisation parameters and different diffuse transposition models are swept, selecting the one that works best for the site. Reductions of the RMSEs between forecasted and real daily AC powers were on average above 70%, dramatically enhancing the clear sky photovoltaic energy prediction. Concerning solar resource irradiance components, the average diffuse fraction of 12 different validation datasets was reduced from an unrealistic 0.517 with the initial turbidity factors extracted from the solar radiation data website (SoDa), to a reasonable 0.224 with the new ones tuned. Moreover, the seasonality of turbidity can be measured using photovoltaics. Whether the tuned factors represent near real on-site aerosols measurements or not is still to be determined.