TY  - THES
AU  - Chen, Shuying
TI  - Variable renewable energy potential estimates based on high-resolution regional atmospheric modelling over southern Africa
VL  - 662
PB  - RWTH Aachen University
VL  - Dissertation
CY  - Jülich
M1  - FZJ-2025-02818
SN  - 978-3-95806-822-3
T2  - Schriften des Forschungszentrums Jülich Reihe Energie & Umwelt / Energy & Environment
SP  - XIII, 141
PY  - 2025
N1  - Dissertation, RWTH Aachen University, 2025
AB  - Africa is the world’s least electrified continent, home to three-quarters of the global population without electricity. Electricity generation in African countries today relies heavily on fossil fuels and hydropower, despite the continent’s abundant potential for the most widely accessible renewable energy sources—wind and solar, as Africa is the sunniest continent in the world and has many windy sites. Africa is also very vulnerable to climate change due to relatively low levels of local socio-economic development. Renewable energy is recognized as an important solution for Africa to address both climate change mitigation and electricity access. Reliable and highly resolved information on Renewable Energy Potential (REP) is imperative to support renewable power plant expansion. However, existing meteorological data sets over Africa used for REP estimates are often characterized by relatively coarse spatial resolution, data gaps in space and time, and general data quality issues. This challenges the reliability and accuracy of existing REP estimates, as well as the modelling of energy systems that include renewable energy. To overcome the existing meteorological data set challenges for renewable energy applications in Africa, the ICOsahedral Nonhydrostatic (ICON) Numerical Weather Prediction (ICONNWP) model in its Limited Area Mode (ICON-LAM) is implemented and run over southern Africa as a prototype for the continent. The ICON model is configured in a hindcast dynamical downscaling setup at a convection-permitting 3.3 km spatial resolution. The simulation time span covers contrasting solar and wind weather years from 2017 to 2019. To assess the suitability of the novel simulations for REP estimates, the simulated hourly 10 m wind speed (sfcWind) and hourly surface solar irradiance (rsds) are extensively evaluated against a large compilation of in-situ observations, satellite, and composite data products. ICON-LAM reproduces the spatial patterns, temporal evolution, the variability, and absolute values of sfcWind sufficiently well, albeit with a slight overestimation and a mean bias (mean error (ME)) of 1.12 m s-1 over land. Likewise the simulated rsds with an ME of 50 W m-2 well resembles the observations. In this work, the simulated 60 m wind speeds (ws60m) from the ICON-LAM simulation and the often-used 31 km-resolution ERA5 reanalysis are also evaluated against measurements at 18 weather masts. The wind power calculated from these simulated wind speed data sets is also compared with measurements at existing wind farms in South Africa. The estimated wind energy potential (WEP) based on ICON-LAM and ERA5 are finally compared using an innovative approach with 1.8 million eligible wind turbine placements over southern Africa. Results show ERA5 underestimates ws60m with a Mean Error (ME) of -1.8 m s-1 (-27%). In contrast, ICON-LAM shows a ME of -0.1 m s-1 (-1.8%), resulting in a much higher average WEP by 48% compared to ERA5. A combined Global Wind Atlas-ERA5 product reduces the ws60m underestimation of ERA5 to -0.3 m s-1 (-4.7%), but shows a similar average WEP compared to ERA5 resulting from the WEP spatial heterogeneity. ICON-LAM also reproduces the observed wind power better than the others, further consolidating the reliability of its derived WEP. Underestimating wind energy yields may hinder the expansion of wind energy, as less economic performance is expected, which underlines the importance of highly resolved meteorological data. Increasing the share of renewable energy in African energy systems is imperative and urgent to address climate change mitigation and access to electricity. This thesis also investigates the impact of the high-resolution ICON-LAM simulations on energy system modelling for southern Africa. An energy system design, encompassing wind energy, solar energy, and battery storage, is derived exemplarily to meet 100% of the local electricity demand, cost-optimized, for each administrative province in southern Africa. Different meteorological data sets, including ICON-LAM as well as the commonly used ERA5 and its variant, are utilized and compared to derive cost-optimized energy systems. The results show significant differences in the wind energy potentials derived from different meteorological data sets, while similar solar energy potentials are found. Cost-optimized energy systems when using ICON-LAM meteorological inputs require less total annual cost (approx. 14%) and battery capacity (approx. 13%) compared to the other energy system solutions using different meteorological input datasets. This suggests that the cost of renewable energy systems may have been overestimated in the past, potentially also hindering its local development. The study further emphasizes the importance of using high-resolution, alternative, atmospheric modelling data sets as a decisive input for energy system modelling. Overall, our results show that the ICON model is able to reproduce the renewable energy related variables and basic atmospheric flows in southern Africa. Compared to other commonly used data sets, the ICON simulations reveal higher wind energy potentials, and cost-optimized energy systems based on these simulations require lower total annual costs and battery capacity. These findings are critical for local renewable energy development, as renewable energy potentials may have long been underestimated and the costs of building renewable energybased energy systems overestimated in southern Africa. Further tuning of physical parameterization schemes specifically for southern Africa may improve the performance of the ICON simulation. Adapting a more sophisticated energy system that includes the real-world power grid and various energy-using sectors may also improve the accuracy of the energy system modelling performed in this study.
LB  - PUB:(DE-HGF)3 ; PUB:(DE-HGF)11
DO  - DOI:10.34734/FZJ-2025-02818
UR  - https://juser.fz-juelich.de/record/1043292
ER  -