Systematic bias in reanalysis-derived solar power profiles & the potential for error propagation in long duration energy storage studies

Abstract

This study consolidates the literature relating to the systematic bias of model-derived irradiance data and further characterises the same for ERA-5 and MERRA-2, two reanalysis datasets that are frequently used to model renewable power in energy system modelling studies. The bias errors yielded when modelling Solar PV generation for solar parks in Australia and Northern Ireland are compared under the optics of typically clear versus typically cloudy respective climates. Evidence is presented that MERRA-2 exhibits a significant global overestimation bias, contrary to some literature that suggests local variation and underestimation in some locations. Based on the trends identified, it is proposed that overestimation bias proliferates in more cloudy climates for both reanalysis datasets. The implications of such bias for studies involving long-duration energy storage with low round-trip efficiency are investigated and an unfortunate mechanism for error propagation is highlighted. A simplified power system consisting of wind, solar, and energy storage is modelled to demonstrate this effect with solar power profiles derived from metered generation, ERA-5, and MERRA-2. It shown that greater systematic bias of MERRA-2, in combination with the opportunity for error propagation through low-round trip efficiency, has the potential to significantly distort the charging & discharging pattern, overall utilization, and total energy requirement of long duration energy storage infrastructure. ERA-5, despite also exhibiting systematic bias, was better able to reproduce the simulated results of the metered generation scenario. Finally, it is recommended that energy modellers using MERRA-2 data to simulate solar PV outputs should now migrate to ERA-5 if reanalysis data is to be used.