A General Approach to Model High-Performance PV Modules for Accurate Energy Yield Simulations

One of the challenges in PV module performance modeling is to obtain a good description of the I-V curves at various temperatures and illuminations, based solely on limited available data. Currently in PVsyst, this is done by calculating the parameters for the one-diode model (1DM) from datasheet information and common assumptions on low-light performance. However, this fails to reproduce high fill factors (FF), typically compromising Voc accuracy. To address this, we use an evolutionary algorithm (EA) which improves parameter determination for the 1DM, extends to more sophisticated models and can use measurement variability as input. This method necessitates solely datasheet information and common low-light assumptions to reproduce IEC 61853-1 measurements. We demonstrate its benefits on both measured PERC modules and synthetic high-FF scenarios. On measured data from a low-FF module, the EA lowers the mean efficiency error in reproducing IEC 61853-1 data compared to the current method. To test the EA on high-FF devices, we generate synthetic data from a reference recombination model informed by solar-cell literature. In this case too, the EA reduces power error and greatly improves Voc reproduction, which is further enhanced when switching from 1DM to the reference model itself.