Multi-junction solar cells electrical characterization by neuronal networks under different irradiance, spectrum and cell temperature

Abstract

Nowadays, HCPV (high concentrator photovoltaics) is largely based on high efficiency MJ (multi-junction) solar cells. Hence, the prediction of the electrical parameters of MJ solar cells is crucial for designing and evaluating the performance of this emerging technology. At the same time, the analytical modelling of the I–V parameters of these devices is complex due to their strong and complex dependence with irradiance, spectrum and cell temperature. In this work, the possibility of predicting the main electrical characteristics of a MJ solar cell by using artificial intelligent techniques is analysed. In particular, three artificial neural network (ANN)-based models were developed: one for simulating the short-circuit current (Isc), one for simulating the open-circuit voltage (Voc) and for simulating the maximum power (Pmax). The models were developed and evaluated with the data of a lattice-matched GaInP/GaInAs/Ge triple-junction operating at a wide range of conditions. Results show that the models accurately estimate the main electrical parameters of a MJ solar cell under different concentrated sunlight, spectral irradiance and cell temperature with a RMSE (root mean square error) lower than 0.5% and a MBE (mean bias error) almost 0%.