The identification and prediction of the daily soiling profiles of a
photovoltaic site is essential to plan the optimal cleaning schedule. In
this article, we analyze and propose various methods to extract and
generate photovoltaic soiling profiles, in order to improve the analysis
and the forecast of the losses. New soiling rate extraction methods are
proposed to reflect the seasonal variability of the soiling rates and,
for this reason, are found to identify the most convenient cleaning day
with the highest accuracy for the investigated sites. Also, we present
an approach that could be used to predict future soiling losses through
the implementation of stochastic weather generation algorithms whose
ability to identify in advance the best cleaning schedule is also
successfully tested. The methods presented in this article can optimize
the operation and maintenance schedule and could make it possible, in
the future, to predict soiling losses through analysis based only on
environmental parameters, such as rainfall and particulate matter,
without the need of long-term soiling data.