Sensitivity of Sub-Hourly Modeling Error to Project Size

High-frequency measurements of solar resource from the Surface Radiation Budget Network (SURFRAD) from stations in NV, MT, SD, MS, PA, IL and CO were down-sampled from 1-minute to 1-hour and used to predict energy yield and sub-hourly modeling error. A Wavelet Variability Model (WVM) incorporating an estimated solar plant layout was used to determine the sub-hourly modeling error dependency for projects ranging in size from 1 MW to 1,000 MW. Additionally, sensitivity to inverter overbuild, DC to AC ratio, average cloud speed, interannual variability and geographic location were evaluated. By incorporating the WVM to smooth the irradiance inputs we found that annual sub-hourly modeling errors exhibited a nearly logarithmic decrease as project size increased. On average, the modeling error decreases quickly for the first 200 MW and begins to asymptote for 200 - 1,000 MW. The magnitude of annual modeling errors was highly influenced by DC/AC ratio, average cloud speed and the interannual variability of the solar resource. The results of this study were implemented to develop a project size dependent sub-hourly modeling error adjustment factor for pre-construction energy assessments.