Simulation of photovoltaic power output using ERA5 reanalysis dataset validated with high-resolution observational measurements

Abstract This study utilizes the ERA5 reanalysis dataset to simulate hourly DC power output using pvlib python (PVLIB) for three years (2018–2020), validated with high-resolution observational measurements from a rooftop photovoltaic (PV) system in Thailand. We have identified the optimal PVLIB configuration to predict the power output from five experiments conducted. Simulations performed better under clear-sky conditions, achieving a correlation exceeding 0.74, root mean square error (RMSE) less than 0.94 kW, and normalized RMSE (nRMSE) less than 18.80%. On the other hand, the performance tended to decline under cloudy-sky, exhibiting a correlation exceeding 0.75, RMSE exceeding 1.0 kW, and nRMSE of more than 94%. Although the performance indicates that ERA5 could be a good substitute data for predicting PV power output, the reliability of these simulations could be improved by assessing additional ERA5 variables and applying bias correction techniques.