This is an old revision of the document!
Since it's currently winter in central Europe, it's the perfect time to evaluate the Unified Clear-Sky Solar output Prediction Model for its feasibility/validity. After bringing the VFCC online, it was a simple task to ship prediction metrics and reference measurements into influxdb and create a couple of dashboards to create meaningful graphs to evaluate its performance very easily.
Yesterday was the first full clear sky day since the beginning of data collection and the prediction results definitely look very promising as we can see on the following dashboard screenshot:
The top right graph shows global solar radiation over a 10 hour period. The green line shows what the UCSSPM predicted as a maximum clear-sky value, the yellow bars show the reference measurement of global solar radiation in W per m² over the same period. The bigger graph at the bottom shows how that translates into the context of photovoltaics (conversion of solar radiation into electricity) and since the output is always a derivative value of global solar radiation the graph looks about the same but the output is measured in W. We can see that we are a long way off the 800-1000 W/m² which are often used for baselining. Of course, it cannot predict the weather but it is an enormous help in planning, scaling and operating solar energy conversion systems in order to have a reliable baseline, so that we know what could be obtained as a maximum at a specific location/time, when there are no clouds.
The reference measurements are sourced from a pyranometer operated by the physics/meteorological department of the Ludwig-Maximilians-University in Munich, who were kind enough to share their data online, which saved us the investment of monetary resources to buy or time to build/calibrate our own pyranometer for now. In order to keep the prediction evaluation results consistent, all other metrics needed by the UCSSPM like temperature, humidity, air pressure etc. are also collected from the same station and fed into the UCSSPM. If time permits, it would be interesting to see if we can make use of our global cloudmap service, to use either historic or live data for an an even more accurate prediction, not just clear-sky baselining.
Long term live reference and UCSSPM prediction metrics are contiously collected and publicly accessible on these VFCC dashboards: