Benchmarking equivalent circuit models for the IV characteristic of bifacial photovoltaic modules

Modelling the current- voltage (IV) characteristic curve of photovoltaic (PV) modules can give valuable insight into the state of health of the PV modules by estimating equivalent circuit parameters, and be used for fault detection in conjunction with IV-based condition monitoring of PV modules, strings and arrays. The most prevalent approach to approximate the IV curve of bifacial solar devices is to model a single diode equivalent circuit model (ECM) with two photo-current sources, representing the front and the back side. This approach is implemented in the main PV modelling software with various methods to parameterize the model. In this work we benchmark the most common ECM models for bifacial devices, using pvlib-python, against module level IV curves measured with an outdoor measuring system. We record IV measurements on eight current-generation bifacial PV modules with PERC-type cells, covering the range of temperature and irradiance conditions over a full year in the Danish climate. The modelling uses datasheet specifications as that is the most widely available input. Results show normalized mean bias errors between - 0.71% and 4.09% and normalized root mean squared errors between 1.62% and 9.76% across P_mp, I_sc and V_oc, with V_oc exhibiting the lowest estimation errors. All models show weaknesses in the irradiance and temperature dependence at I_sc. The CEC and PVsyst model have an improved temperature and irradiance dependency at V_oc compared to the DeSoto model, which performed better at estimating P_mp. While the models characterize the V_mp/V_oc ratio well, they fail to cover the wide range of I_mp/I_sc ratios seen in measurements. We found the generic datasheet parameters used for model parametrization to be overestimated compared to flash tests, which contributed to the modelling error. Using non-module specific inputs also meant that results vary significantly for different modules of the same type. A monofacial module showed on par or worse modelling results compared to bifacial modules, validating the modelling approach of a combined effective irradiance for the front and back side.