Estimation of the Performance of Photovoltaic Cells by Means of an Adaptative Neural Fuzzy Inference Model

Abstract

This paper presents an Adaptive Neuro-fuzzy Inference System capable of predicting the output power of photovoltaic cells using their electroluminescence image and their IV curve. The input consists of 3 different features: the number of black pixels, grey pixels and white pixels. ANFIS combines the learning capabilities of Artificial Neural Networks with the comprehensible rules of Fuzzy Logic, being optimal for this problem, as demonstrated by the metrics of MAE of 0.064 and MSE of 0.009, which are better than the performance of other tested methods such as Support Vector Machines or Linear Regressor.