Reanalysis datasets have been extensively used to estimate the power generation potential of solar photovoltaic (PV) systems owing to their global coverage, inclusion of multiple variables, and high temporal resolution. However, the spatial resolution of these datasets is substantially coarser than the typical area of PV farms, which can lead to potential inaccuracies in power generation estimates. Despite this limitation, the associated errors have not been thoroughly investigated. In this study, we comprehensively assess the misestimations in power generation potential arising from the application of three widely used reanalysis datasets—ERA5, ERA5-Land and MERRA2—using observational data from twelve PV farms in China. Moreover, the key drivers and mechanisms underlying these inaccuracies are revealed. We find that the relative errors in annual power generation potential estimates were approximately −17%, −12% and 7% for ERA5, ERA5-Land and MERRA2, respectively. Notably, these errors escalated with increasing temporal resolution of estimation; at the hourly scale, they were amplified by approximately twofold and threefold for ERA5/ERA5-Land and MERRA2, respectively. Mechanistic analysis revealed a significant negative correlation between estimation errors and near-surface air temperature, suggesting that cloud conditions and terrain may be the primary drivers contributing to these inaccuracies. The results of our study underscore the importance of accounting for these errors when using reanalysis data for PV power generation potential estimation, particularly at monthly and hourly scales. Policymakers and researchers can utilize our results to quantify the uncertainties in PV power generation calculations and their subsequent applications.

Read original article

Read original article