For the last years, PV (photovoltaic) generation has been having an important impact on LV (low voltage) and MV (medium voltage) grids in Spain, due to the increasing number of installations. As power monitoring of the generation is in most cases not required, utilities are forced to make assumptions on PV power generation in order to perform network planning studies for both peak demand and contingency analysis. These assumptions increase errors committed during the analysis, as the number of PV installations increases. This paper presents a methodology for estimating PV active power generation values for planning purposes in MV and LV power systems, from historic generation data, based on the development of correlation models. This methodology is applied to three different examples, using predictors based on real registered data. The methodology was also applied in a typical grid study and it's error was determined. (C) 2013 Elsevier Ltd. All rights reserved.