Sensitivity analysis of PV power simulations for different temporal resolutions and spatial aggregation levels

The power generation data from distributed PV systems is of high importance for different stakeholders and is useful in numerous applications. However, PV power measurements are often only available for a small fraction of the total number of systems. As a result, the power generation of those remaining PV systems is estimated using upscaling approaches or satellite-based simulations. This paper evaluates the impact that the simulation environment has on the accuracy of power upscaling in four different scenarios for different temporal resolutions and spatial aggregation levels (portfolio sizes). Using three methodological steps, we show that a standard scenario exhibits improved accuracy with either a higher resolution of satellite images, consideration of module orientations or a calibration. The validation basis consists of 728 PV systems located within four districts in Germany, with measured power values recorded between 2016 and 2017. The successive refinements in the different scenarios show significant improvements in the rRMSE and rMBE. A lower temporal resolution further improves the rRMSE but no notable impact can be observed for the rMBE. An increasing portfolio size leads to continuous improvement in both error metrics. For benchmarking purposes, these error metrics are further compared to the operational nowcasting accuracy of larger PV portfolios from grid operators and very similar characteristics are observed.