This paper presents a novel approach for tracking the maximum power point of photovoltaic (PV) systems so as to extract maximum available power from PV modules. Unlike conventional methods, a very fast tracking response with virtually no steady state oscillations is able to obtain in tracking the maximum power point. To apply the proposed method, firstly, output voltages, output currents under different conditions and temperatures of a PV module are collected for the fitting of environmental invariant nonlinear model for the PV system. Orthogonal least squares estimation algorithm coupled with the forward searching algorithm is applied to sort through all possible candidate terms resulted from the expansion of a polynomial model and to come up with a parsimonious model for the PV system. It is not necessary to test all PV modules as the resultant model is valid for other modules. The power delivered by the PV system can be derived from the fitted model and the maximum power point for the PV system at any working conditions can be obtained from the fitted model. Consequently, rapid maximum power point tracking could be achieved. Experimental results are included to demonstrate the effectiveness of the fitted model in maximum power point tracking. Crown Copyright (C) 2013 Published by Elsevier Ltd. All rights reserved.