Accurate sizing is one of the most important aspects to take into consideration when designing a stand-alone photovoltaic system (SAPV). Various methods, which differ in terms of their simplicity or reliability, have been developed for this purpose. Analytical methods, which seek functional relationships between variables of interest to the sizing problem, are one of these approaches. A series of rational considerations are presented in this paper with the aim of shedding light upon the basic principles and results of various sizing methods proposed by different authors. These considerations set the basis for a new analytical method that has been designed for systems with variable monthly energy demands. Following previous approaches, the method proposed is based on the concept of loss of load probability (LLP)-a parameter that is used to characterize system design. The method includes information on the standard deviation of loss of load probability (sigma(LLp)) and on two new parameters: annual number of system failures (f) and standard deviation of annual number of failures (sigma(f)). The method proves useful for sizing a PV system in a reliable manner and serves to explain the discrepancies found in the research on systems with LLP < 10(-2). We demonstrate that reliability depends not only on the sizing variables and on the distribution function of solar radiation, but on the minimum value as well, which in a given location and with a monthly average clearness index, achieves total solar radiation on the receiver surface. (c) 2007 Elsevier Ltd. All rights reserved.