Assessment of Internal Power Losses in Photovoltaic Cells Using an Adaptive Neuro-Fuzzy Inference System Based on Electroluminescence

Series resistance (Rs) is a key parameter that limits photovoltaic cell performance; however, its conventional estimation from current–voltage measurements requires electrical contact and controlled testing conditions. In this study, a nondestructive image-based methodology is proposed to estimate Rs in crystalline silicon photovoltaic cells using electroluminescence images and a Sugeno-type adaptive neuro-fuzzy inference system. A dataset of 666 electroluminescence images was processed using normalized grayscale histogram descriptors, and reference Rs values were obtained from I-V characterization. Three global radiometric descriptors corresponding to low-, medium-, and high-intensity pixel fractions were used as model inputs. The ANFIS model achieved high predictive performance, with a testing RMSE of 0.0065 Ω and an R2 value of >0.98. These results indicate that the global EL intensity distributions contain information related to resistive losses. However, further validation under different acquisition conditions and using independent datasets is required before field-scale deployment. The proposed approach provides a compact and interpretable framework for rapid photovoltaic cell screening based on electroluminescence imaging, complementing conventional electrical characterization under controlled laboratory conditions.