Models based on daily energy balance (or long‐term models) have been widely used as a tool in the stand‐alone photovoltaic (PV) system sizing, mainly with the purpose of obtaining analytical expressions of the relation between the generator size and the storage capacity of the battery. The system can then be designed to meet the reliability requirements of the specific case. However, such models represent the complex operation of a stand‐alone system in an oversimplified way. There is little research so far on the reliability and improvement of such models. Validation and possible modification of a long‐term system model requires comparison of the simulated state of charge (SOC) of the battery with that obtained from an experimental system. In this work, experimental data from a 6‐month operation of a basic stand‐alone PV system have been analysed and compared with modelling results. One obvious improvement that could be applied to the long‐term system model is to account for a charging efficiency of the battery, and this possibility is examined in the present work. However, comparison with the modelling results shows that the data cannot be fitted by simply taking into account battery inefficiency. A method to account for system memory effects in the increase of the battery SOC, imposed by the operation of the regulator, is necessary to accurately model the macroscopic diurnal charging/discharging process. Copyright © 2012 John Wiley & Sons, Ltd.