Forecasting Solar Power and Irradiance – Lessons from Real-World Experiences
Abstract
MDA Information Systems, LLC developed a solar irradiance and power forecasting system based on a first principles science foundation employing high-quality scientific datasets such as AERONET and SURFRAD and utilizing the REST2 clear sky model as an underlying basis for the full-sky forecast. Real-time inputs include a diverse multi-model ensemble of numerical weather prediction (NWP) forecasts, ground-based solar monitoring observations and proprietary observations of solar power from client sites, and visible satellite imagery. Forecasts were made for challenging locations where daytime cumulus clouds and occasional storm systems passing through resulted in variability on time scales of minutes, hours, and days. This paper focuses on lessons learned from our experience with real-world data and real-world power and irradiance forecasts. Topics include quality control of irradiance and power observations, sub-hourly variability and inverter-limited sites, tracking angles for single-axis trackers, and situational bias of NWP forecasts.
Keywords
Data, Ensemble, Forecast, Irradiance, Model, Numerical weather prediction, Observations, Power, Quality control, Real-time, Tracking, Variability