The aim of this paper was to study total solar radiation using a Markov cha
in, with a view to setting up a solar data generator. This meteorological p
arameter was transformed into another, called the clearness index, and was
classified into 20 states. The various transition probabilities between eac
h state were computed. Markov chain properties were investigated and confir
m the choice of a first order. The occurrence probability of same state seq
uences was then studied. It was shown that it can be modelled using a proba
bility law, called the shifted negative binomial (SNB) distribution. Finall
y, an hourly solar data generator was set up based on these two models, and
the real and simulated series from a probability point of view were compar
ed. This confirmed a good agreement of the proposed generator to obtain a t
ypical meteorological year for solar data. Copyright (C) 2000 Royal Meteoro
logical Society.