A scalable method for extracting soiling rates from PV production data

We present a method for analyzing time series production data from photovoltaic systems to extract the rate at which energy yield is affected by the accumulation of dust, dirt, and other forms of soiling. We describe an approach that is based on prevailing methods, which consider the change in energy production during dry periods. The method described here builds upon these methods by considering a statistical sample of soiling intervals from each site under consideration and utilizing the robust Theil-Sen estimator for slope extraction from these intervals. The method enables straightforward application to a large number of sites with minimal parameterization or data-filtering requirements. Furthermore, it enables statistical confidence intervals and comparisons between sites.