The estimation of the regional photovoltaic (PV) power generation is an important step prior to the forecast of the PV power generation and its integration into the energy supply system. The large majority of PV plants not being measured in Germany, the total PV power generated in a region is commonly estimated by upscaling the power production of a set of reference PV plants to the entireness of the plants installed in the considered area. A given uncertainty can be expected in the estimation of the power generation of a PV plant with the upscaling method when the reference plants used have different configurations or weather conditions. To gain better insight into the performance of the upscaling method, its error has been analysed using power measurements of a set of 366 PV plants. The analysis allows an understanding of the mechanisms underlying the uncertainty of the upscaling method and quantifies its error for the test study considered. In the case study analysed, it could be shown that the quarter hourly RMSE1 value decreases with an increasing number of reference plants and a decreasing number of un-metered plants. It could also be shown that even for a large number of reference plants, a variation of the RMSE between 0.01 and 0.025 kW/kW(p) can be observed, depending on the choice of the reference plants. It is shown that the average distance between a reference and unknown plant constitutes a good indicator of the performance of a set of reference plants, but that the match between the characteristics of the reference and unknown plant also plays an important role, which could not be quantified with the available dataset. (C) 2016 Elsevier Ltd. All rights reserved.