Development, Implementation, and Integration of a Holistic Solar Forecasting System for California

Errors in forecasting the development and breakup of fog and low coastal clouds are an important contributor to errors in forecasts of California’s solar generation. To improve solar generation forecasting capabilities, this project set up and performed a series of experiments using the Weather Research and Forecasting numerical weather prediction model. These experiments tested the sensitivity of the model’s ability to forecast foggy conditions using a range of parameterization choices. A measurement program involving targeted use of groundbased atmospheric boundary layer sensors was successfully completed from December 2017 through March 2019 to support improvements in predicting fog and stratus cloud (low level cloud) dissipation in high solar penetration regions of California.

Analysis of power grid operating data showed no simple correlation between solar forecasting errors and electric grid operation metrics. Production cost modeling simulation analysis showed that with 20 percent solar forecast accuracy improvement in day-ahead forecasting, the fuel cost and startup cost of thermal units in the system could be reduced by between 0.06 percent and 3 percent, depending on test systems. These results can help to guide users and producers of solar generation forecasts towards prioritizing and properly valuing improved forecasting technologies. Finally, the impacts of improved solar power forecasting and battery storage resources to the system were compared. In the modified California system, the economic value of 20 percent solar power forecasting improvement is similar to that of a 25 MW battery storage resource.