This article presents a method to estimate the parameters of a photovoltaic cell model in its equivalent circuit of a single-diode using artificial neural networks; more specifically, the multilayer perceptron concept is used. The data required to estimate the parameters are based solely on the information available in the manufacturer's data sheet. The neural network has two hidden layers and the selected training method was Bayesian regularization. The training data were produced synthetically to ensure that a large part of the range of possible parameter values is covered. The network performance is validated with a data set not included in the training set, making the comparison with a recognised method from the literature, and even more, with experimental data obtained from a real photovoltaic panel. The main advantage of this method is that, once the network is trained, the parameters returned by the network will always be unique for each input provided, which is not the case with other artificial intelligence methods such as genetic algorithms or differential evolution. In addition, this method can be directly applied to other similar problems, only by modifying the inputs and outputs of the network.