In this work, a radial basis function network model of solar cells is developed and validated against measured data. A grid-interpolation based data pre-processor is developed to prepare the training data from non-uniformly distributed measured data obtained via normal operation of the solar cells. Simulation results show that the pre-processor facilitates training, and that the resulting model is accurate under conditions sufficiently represented by the training data. The model matches measured data more accurately compared to conventional solar cell models. Copyright (C) 2003 John Wiley Sons, Ltd.