To estimate the accumulated PV power feed-in of an region, measurements of reference PV systems can be used. Since the data of a perfectly representative ensemble of systems is hard to obtain, statistical post-processing methods can be used to correct resulting estimation errors. In this paper the Analog Sampling algorithm is introduced to correct diurnal error patterns by taking the solar angle into account. It is a local regression method that uses the Analog Ensemble metric for fitting linear models on localized subsets of the data. The algorithm is evaluated on the basis of measurements from meteocontrol GmbH and real feed-in values from the Tennet control area. It can be shown that it is able to correct errors resulting from the use of a predominantly utility scale reference ensemble.