Leveraging Green Ammonia for Resilient and Cost-Competitive Islanded Electricity Generation from Hybrid Solar Photovoltaic–Wind Farms: A Case Study in South Africa

Hybrid solar photovoltaic (PV) and wind generation in combination with green ammonia as a seasonal energy storage vector offers an excellent opportunity to decrease the levelized cost of electricity (LCOE). In this work, an analysis is performed to find the most cost-effective configuration of power-to-ammonia-to-power (P2A2P). In P2A2P, wind and solar resources are combined with energy storage to design a resilient electricity grid. For daily generation, batteries are utilized for energy storage, whereas ammonia is employed to cope with seasonal fluctuations. The costs of energy storage capacity have a significant influence on the LCOE. Therefore, this work studies the effect of solar/wind hybrid generation systems and energy storage capacity on the LCOE. A base case of the region of De Aar in South Africa was selected because this inland location has excellent wind and solar resources. The optimized battolyzer and Haber–Bosch design capacity led to an overall load factor of 20–30%. At a 30% load factor, a hybrid system with 37% wind-based and 63% solar-based energy generation capacity was the most cost-effective configuration, resulting in a LCOE of 0.15 USD/kWh at a 5% annual discount rate. In an optimistic scenario for PV costs, the LCOE achieved is essentially unaltered (0.14 USD/kWh), while the contribution of wind and PV changes to 25 and 75%, respectively. This analysis indicates that appropriate designing of hybrid energy solutions will play a key role in determining the final energy storage capacities needed to reduce the LCOE. While these costs for LCOE are above those reported for coal-powered electricity in South Africa (e.g., 0.072 USD/kWh for businesses and 0.151 USD/kWh for households), a carbon tax of 50 USD/ton of CO2 can increase these costs to 0.102 and 0.191 USD/kWh, rendering a more promising outlook for the P2A2P concept.


General methodology:
Figure 1 illustrates a process scheme that starts with the demineralization of brackish water to subsequently utilize for electrolysis of water.For the demineralization, a Zero-Liquid-Discharge (ZLD) method is utilized as proposed in 1 .The ZLD process combines High Rejection Reverse Osmosis (HRRO), together with Low Rejection Reverse Osmosis (LRRO) as described in 2 , and Mechanical Vapor Compression (MVC).In the next step, battolysers are utilized to store energy on a short term scale, and to produce hydrogen.The hydrogen gas is subsequently utilized to thermal-catalytically produce ammonia in the AE-HB process.The nitrogen for ammonia production is obtained using Pressure Swing Adsorption (PSA), due to its high flexibility and reasonable energy consumption [53] 3 .For the ammonia storage facility, ammonia is stored in large scale refrigerated storage tanks at atmospheric pressure and -33 o C. For large-scale ammonia storage, refrigerated storage at atmospheric pressure is the most cost effective and common storage method, alternatives are high pressure storage at room temperature and storage is salts 4 .
To convert ammonia back into electricity, several process options are possible.The two main options are utilizing direct fuel cells, or fuel cells with thermal energy recovery.From the given options, utilizing direct ammonia fuel cells utilizing Solid Oxide Fuel Cells are the preferred option 5 .Heat recovery technologies are not preferred due to their poor electrical efficiency gain, together with significant increase in CapEx 6 .The efficiency of the SOFC is assumed to be 55%LHV in this work 7 .The utilization of direct ammonia fuel cells is possible as the electrocatalyst in SOFC is Ni-based and does not require pre-dissociation as Ni-based electrolysis is able to dissociate NH3 at temperatures present within SOFC 8 .This makes direct fuel cell technologies attractive, as external ammonia cracking can result in heat and thus efficiency losses 7 .
The specific energy consumption (SEC) of ammonia production was 12.0 kWh/kg NH3, as calculated in 9 (see table 1).Modern ammonia plants based upon natural gas or coal have an energy consumption of 7.8 and 10.6 kWh/kg NH3, respectively.Utilizing SOFC-H with an electrical efficiency of 55%LHV, an RTE of 24% is found for seasonal storage.Although this is deemed low, it is higher than the well-to-power (W2W) efficiency of fossil based technologies, which often range in 16-20% 10 .With () ℎ,ℎ defined as the difference in energy generation and energy demand in a given hour of the year.
Subsequently, the amount of energy that is lost in a year, and should be compensated using ammonia is calculated according to the following equation: The average of the three years is taken as the additional amount of energy that is required to be generated by ammonia.On the seasonal scale, a similar calculation is performed to find how electrolysis and Haber-Bosch under sizing effects ammonia production.With the two loss factors combined, one can calculate the increase in energy generation capacity required to calculate the ammonia quantity, and ammonia generation capacity required to design an islanded electricity system.
For the weather data, data from 2019, 2020, and 2021 has been utilized from the Copernicus database with 1 data point per hour, resulting in a string of.Specifically, the solar radiation (total sky direct solar radiation at surface) and the wind speed (U-and V-direction wind speed at 100 m hub height) have been retried.This results in a total of 26.304 x 3 = 78.912data points that are used as an input in this work.For the demand pattern, data is obtained from the previous 5 years, up to the middle of 2022.Due to the non-complete data from 2022, this year is disregarded.Furthermore, the demand data from 2020 is disrupted due to the COVID lockdown, and the demand from before the lockdown is deemed to be outdated.Therefore, data from the complete 2021 year is taken as an input for demand patterns.For wind to energy output correlations, the Vestas V82 1.65 MW wind turbine 18 is modelled, whereas a linear correlation between solar radiation and PV output at a 15% efficiency is assumed for the PV generation capacity 19 .For wind and solar CapEx estimated, data from 20 and 21 are utilized, respectively.Cost estimates for PV vary widely in literature, therefore a good case scenario from 22 is utilized in this work as well.For OpEx, the wind turbine OpEx costs from 23 are scaled linearly with the CapEx from the South Africa case, as compared to the CapEx from the given source.For PV OpEx, a constant 1.1% annually from the CapEx is used 22 .

Solar and wind patterns
From the Copernicus database, solar and wind data has been retrieved from 2019, 2020, and 2021.The data has been processed and are visualized in figure S1.From figure S1, it shows that solar energy is highly cyclic, on an hourly scale as well as on a seasonal scale.Wind energy, on the contrary, is more stable throughout a day and the year.On a daily basis, solar PV and wind are complementary, noting the relatively low wind speeds during the day, when solar PV has a peak capacity.On a monthly basis, solar PV and wind are less complementary.

Sensitivity analysis
To understand the influence of energy generation equipment CapEx and design capacity, a sensitivity analysis is performed.Due to the large differences in costs estimates for PV, two scenarios are plotted.These are the base case scenario of 961 USD/kW, and the good case scenario of 618 USD/kWh.LCOE results are plotted for relative battolyser + HB design capacities of 10%, 20%, 30%, and 100%, and varied from a grid of 100% wind / 0% solar, to 0% wind / 100% solar and are illustrated in figure S2.Curve smoothing has been applied to the graphs to diminish the noise in the graph, which is the result of several iterative steps in the script.This is however, not performed for the graphs at 10% design capacities, as an incorrect trend would be suggested at low (<5%) percentages of solar in the grid.It becomes clear that although more solar is present in the system as compared to wind energy, the main cost contributor to the LCOE is the wind turbines.This is because the CapEx per peak kW of wind energy is twice as high as compared to solar (1877 USD/kW versus 961 USD/kW), whereas operational costs for wind turbines is almost 4 times higher (2.7% of CapEx/year versus 1.1% of CapEx/year) and additional large maintenance is required for wind turbines after 10 and 15 years costing 17% and 13% of the CapEx, respectively.At last, it is important to state that the costs of wind turbines, solar parks, and battolysers can influence the optimal grid configuration, whereas the costs of the SOFC-H do not impact this optimum.This is because, regardless of the price, the SOFC-H sizing is based upon peak energy demand assuming that no energy is available from wind turbines, PV panels, or the battolyser.

Figure S1 :
Figure S1: a) average hourly solar radiation, b) average hourly wind speed, c) average monthly solar radiation, d) average monthly wind speed in South Africa.

Figure S3 :
Figure S3: sensitivity analysis of process equipment costs on LCOE.
2. Energy efficiency estimationFor a full capacity system, the model is designed so that sufficient electricity is charged to cope with daily energy cycli, assuming an averaged cyclus of generation and demand, and a battolysers RTE of 85%.With reducing battolyser capacity, the required amount of short term energy storage will reduce as well.Because of this, ammonia is required to compensate for short term energy shortages.The amount of short-term ammonia losses as a fraction of the total charging rate is calculated for each hour in 2019 to 2021 according to: () , = max (( () ℎ,ℎ − () ℎ,ℎ () ℎ,ℎ ) , 0)