Evaluation of Wind and Solar Power Generation over South Africa according to Different Data Sources

Abstract

Reanalysis data and regional downscaling using atmospheric models have become integral tools for assessing wind and solar energy potential in renewable energy simulations. This study leverages ERA5 reanalysis and the high-resolution ICON in Limited Area Mode (ICON-LAM) simulated dataset, along with the Renewable Energy Simulation toolkit (RESKit), to evaluate wind speed and solar radiation, followed by the evaluation of their conversion to power using the RESKit model. South Africa is chosen as our study domain for its current national wide poor electricity supply and the availability of observation data. The focus time period spans a duration of three years, ranging from 2017 to 2019. In the first step, observation wind speeds collected from weather masts and observed solar radiation collected from Southern African Science Service Centre for Climate Change and Adaptive Land Management (SASSCAL) and Trans-African Hydro-Meteorological Observatory (TAHMO) stations are compared to modelled wind speeds and solar radiation. In the second step, observation wind and solar power generation collected from the Renewable Energy Data Information Service (REDIS) are compared to modelled wind and solar power generation. This comparison is conducted through diverse indicators, including Pearson correlation (r), root mean square error (RMSE), mean error (ME), coefficient of determination (R-squared or R2), mean absolute error (MAE), and Perkins skill score (PSS). Our findings underscore the superiority of ICON-LAM over ERA5 in terms of ME, MAE, R-squared, RMSE, r and PSS for wind speed assessment. For solar radiation analysis, ICON-LAM outperforms ERA5 in terms of PSS, MAE, and RMSE. Shifting focus to wind power estimation, ERA5 shows better performance than ICON-LAM in ME, R-squared, RMSE, Pearson correlation, and PSS. Similarly, in solar power estimation, ERA5 excels over ICON-LAM in MAE, ME, RMSE, r, and PSS. It's crucial to highlight that the power generation comparison has been carried out using observation data that is significantly aggregated, resulting in notable uncertainty that may constrain the performance evaluation of the high-resolution ICON-LAM. This study not only underscores the significance of advanced datasets and tools but also sheds light on their nuanced performance in assessing renewable energy potential across various metrics.