The Challenges of Solar Forecasting: Reducing the Cost of Solar Power Through Research

Carlos Coimbra [UC Merced]
Abstract:

The Solar Forecasting Laboratory at the University of California Merced has collected over 15 months of high quality horizontal and direct normal irradiance measurements at different wavelengths (UV, IR and visible) with the primary objective of developing, calibrating and benchmarking novel and more accurate forecasting models for solar irradiance at the ground level. Without effective forecasting methodologies, neither solar nor wind power plants cannot be effectively connected to the power grid, which presents a major obstacle for high-penetration utilization of intermittent sources. We will discuss different strategies for solar forecasting for both short- and long-term, high spatial and temporal resolution, and forecasting for different solar applications. We will also present a hybrid (GA/ANN) SSL (Stochastic, Self-Learning) methodology that yields very promising results for modeling and forecasting the solar resource, as well as some recent results of the application of our forecasting methodology to model the power output of UC Merceds 1 MW solar farm (SunPower 1-axis TPV).