Artificial neural network approaches for modeling absorption spectrum of nanowire solar cells

Abstract

The absorption spectrum of the silicon nanowire solar cells has been modeled using artificial neural networks. Multi-layer perceptrons (MLP) have been proposed as a precise and promising neural network to model the absorption of the nanowire solar cells with a very high accuracy. The MLP results have also been compared to the radial basis function (RBF) method. The proposed algorithm could successfully model the effect of the geometrical parameters of the nanowire array such as the nanowire diameter and pitch on the absorption spectrum. Finite difference time domain calculations show that the MLP network has tracked the sharp variations of the absorption spectrum more precisely. An optimum number of the neurons and epochs for MLP network have been obtained (8 neurons and 100 epochs). The results show that the optimized MLP network achieved a lower consuming time and error than RBF network. The calculated mean square error for the MLP network has been 0.0003 which verifies the high efficiency and accuracy of the proposed neural network model.