Genetic algorithm-trained radial basis function neural networks for modelling photovoltaic panels

Radial basis function neural networks (RBFNs) can be applied to model the I-V characteristics and maximum power points (MPPs) of photovoltaic (PV) panels. The key issue for training an RBFN lies in determining the number of radial basis functions (RBFs) in the hidden layer. This paper presents a genetic algorithms-based RBFN training scheme to search for the optimal number of RBFs using only the input samples of a PV panel. The performance of the trained RBFN is comparable with that of the conventional model and the training algorithm is computationally efficient. The trained RBFNs have been applied to predict MPPs of two different practical PV panels. The results obtained are accurate enough for applying the models to control the PV systems for tracking the optimal power points.