Data requirements for calibration of photovoltaic system models using monitored system data

Performance models for photovoltaic modules are traditionally calibrated from measurements of module output under carefully controlled indoor or outdoor conditions. Recently Sandia National Laboratories has published methods to obtain performance model parameters from measured output of modules outdoors on fixed tilt racks. Here we investigate the dependence of a model´s prediction accuracy on the amount of data used to calibrate the model´s parameters. We find that performance models with acceptable accuracy can be obtained from relatively limited data sets.