734
Forecasting Electric Generation from Solar Energy

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 5 February 2014
Hall C3 (The Georgia World Congress Center )
Stephen D. Jascourt, MDA Information Systems, LLC, Gaithersburg, MD; and D. Kirk-Davidoff, C. Cassidy, and T. Hartman

Handout (828.8 kB)

The rapid increase in solar power on the grid is driving a need for utilities, independent system operators, and others to have accurate forecasts of the amount of solar power generated. In response, MDA Information Systems, LLC has developed a solar power forecasting system. Last year at this meeting, we reported on the MDA state-of-the-science irradiance forecasting system which utilizes the REST2 clear sky model, AERONET aerosol observations, and other public sources as well as private site data. Ultimately, many users of the information are primarily interested in the electric power generated. A sharper focus on predicting the generation output from real, not ideal, PV systems using real, not perfect, data is crucial. Therefore, the emphasis in this presentation will be on the electric power generation and lessons learned through dealing with site data, as well as discussing other applications of irradiance data to the utility industry.

Some of the lessons learned include:

* Addressing the efficiency of electric generation as a function of irradiance and other variables

* Site data reveals key differences from manufacturer specs. Site data are key to making a good power forecast

* Sun-tracking systems may malfunction, affecting the angle of the solar panels and thus the energy available to them and resulting power output. The actual angles can be reliably calculated under clear sky conditions or over several days when there are enough sampled clear-sky times

* Site data requires quality control, including removing erroneous readings that are within a physically plausible range (such as when the instrument or recorder gets stuck for a short period)

* Care must be taken to not include times of accumulating snow in datasets used for training or developing algorithms

As we develop further experience before the conference, we will learn more lessons which can help raise the bar on solar energy forecasting. Additionally, MDA is participating in the large NCAR-led DOE-supported solar forecasting initiative and we may have other developments stemming from that collaboration to report on at the conference.