Accuracy of energy assessments in utility scale PV power plant using PlantPredict

The prediction methodology of First Solar's proprietary modeling tool PlantPredict was validated in two ways. First, 51 equivalent simulations were run in PVsyst and PlantPredict, indicating a mean energy yield difference of 0.13% ± 0.52% (1-sigma). Second, in-depth actual-expected analyses were performed on 20 utility scale systems with nearly 1 GWDC of First Solar CdTe modules installed. On average, PlantPredict underpredicted energy by 0.41% ± 2.01% (1-sigma) using First Solar's energy prediction guidance. Use of the Perez transposition model rather than the Hay & Davies model increased plane of array transposition error by 1.33%. The Hayes advanced thermal model reduced the hourly module temperature RMSE by 43%, but the annual module temperature model accuracy did not change significantly. Finally, computing spectral shift on an hourly basis was shown to not significantly increase model accuracy but reduced the standard deviation of daily model error by 4.25%.