Assessment of a new simulation approach for estimating PV output variability from satellite imagery

Uncertainty about the impacts of very large PV plants on utility operations is a barrier to building such plants. Output variability is of particular concern due to its potential impact on grid reliability and operations. Quantifying the impacts of proposed plants generally relies on simulating plant performance without the benefit of irradiance measurements at the proposed plant locations. Accordingly, methods for simulating irradiance are of great interest. Sandia National Laboratories (SNL) has developed a relatively simple method to estimate one-minute power output from fleets of large PV plants. The method relies on generating artificial time series of irradiance from hourly satellite irradiance estimates (available nearly everywhere) and one-minute ground measurements of irradiance from analogue sites in the region [1]. When applied to the southern Nevada region, the resulting time series preserve basic statistics for irradiance and for changes in irradiance, and exhibit characteristics evident in measured irradiance from other regions. Here we investigate the dependence of the simulation on the available ground measurements. If ground measurements capture generic irradiance patterns, then the method could be applied to regions other than where the measurements were made. We use ground measurements from southern Nevada and Tennessee to simulate irradiance for a site in Florida, where we use hourly averages of ground-measured irradiance in lieu of satellite estimates. The simulation results compare poorly with ground observations of irradiance at a one-minute time scale. We conclude that our simulation method is not general in the sense that it appears to require irradiance measurements within the region of interest.