Performance of Separation Models to Predict Direct Irradiance at High Frequency: Validation over Arid Areas
Abstract
A comprehensive study of the performance of 36 separation models selected from the literature is presented here, using high-quality 1-min data of global horizontal irradiance (GHI) and direct normal irradiance (DNI), toward the evaluation of the uncertainty in GHI-derived DNI. A detailed performance assessment is conducted from 9 stations over arid or desert areas of 5 continents, where the solar resource is high and solar systems have great potential. To evaluate the performance of each model, three summary statistics are calculated. The random errors are found significant, even though the test stations have only low cloudiness compared to temperate climates. For some models, the errors are exacerbated by cloud enhancement effects. The uncertainty in the predicted DNI appears highly dependent on the local radiation climate, the specific separation model, and the number of predictors used. The two Perez models, which both use a variability predictor, are most generally those generating the best predictions, although they conversely have more bias than simpler models, and may occasionally generate spurious results.
Keywords
Direct-diffuse separation, Dni, Irradiance variability, Cloud enhancement, Validation.