It is generally agreed that nuclear power comes with an inherent risk. What is hotly debated is whether or not there are benefits which outweigh that risk. When it comes to the Koeberg Nuclear Power Plant, Eskom have had the approach that since Koeberg produces electricity at a cost of around R0.45 per kWh, it is a ‘no-brainer’ to spend whatever is needed on a refurbishment to extend the life of the plant beyond the end of its design life in 2024.
A recently released modelling report estimates that doing that refurbishment would result in additional costs of R50bn “from a combination of the actual costs of life extension, additional open cycle gas turbine (OCGT) usage, large key-customer curtailment requests/instructions, and scheduled load-shedding.” It will also increase carbon emissions due to the additional use of OCGTs.
Independent energy analyst Clyde Mallinson has been working on modelling the South African electricty landscape since 2017. Using hourly historical data of demand and supply, his model allows ‘what if’ scenarios to be rapidly analysed.
For example, given a year of solar power generation and the installed capacity of solar at the time, it is reasonable that a percentage increase in installedsolarsolar capacity will result in the same percentage increase in solar energy availability. This is because the variability inherent in solar power is already ‘baked in’ to the data.
In discussions with Peter Becker of Koeberg Alert Alliance, the possibility arose of applying this national model to the more narrow question of studying the impact of extending the life of Koeberg. A day later Mallinson had some preliminary results which he found unexpected and surprising.
His model compares two scenarios: firstly taking each unit offline in 2022 for about 6 months in order to replace and refurbish multiple components of each reactor unit, as Eskom is planning on doing. The second scenario is to limit outages to those needed for refuelling, but apart from that to run them as much as possible until July 2024, and then to shut them down permanently as initially planned. The time between now and 2024 can be used to build new renewable energy and storage capacity to take over from Koeberg when it is retired. Mallinson tabulates the results, and analyses the critical differences between the two.
Each scenario results in an hourly model for the year, one week of which in 2022 looks something like the following:
The red curve is demand, which is ‘filled in’ with the generation sources available in each hour. Coal comprises the black area, and the unfilled white gaps, for example on Thursday above, indicate a shortage of supply, which means load shedding or voluntary curtailment is required to avoid system failure. The orange areas above the red curve indicate supply exceeds demand, and that excess supply is put into energy storage, such as pumped storage.
The above shows the case where both Koeberg units are kept running throughout 2022. The next graph (Figure 2 below) shows the result if each unit is offline for refurbishment for six months of the year, thereby halving the contribution from Koeberg. The increase in load shedding is clear – six days include some load shedding as opposed to three days in the above graph.
Figure 2: Koeberg re-furbished in 2 x 6-month steps in 2022
The final graph (Figure 3 below) shows a scenario in 2025 in which Koeberg has been shut down in 2024, and replaced by renewables and storage. There is no more load shedding, and there are times when far more energy is produced than is needed. This is ‘free’, or as Mallinson likes to call it, ‘super power’, which could be used to charge electric vehicles, or sold at a very low rate to power industries which can use variable power, such as plastic recycling plants.
For 2022, the model shows the following consequences if the long term outages and refurbishment goes ahead.
More expensive electricity
Doing the refurbishment to extend the life of Koeberg will result in an increase of 6c per kWh in the cost of generating electricity. The reason for this is the need to use peaking power plants to fill in during the long outages.
Since peaking power plants use fossil fuel this also means more carbon emissions, which the model estimates to be 3Mt of carbon. The model does not include the impact of other emissions such as sulphur compounds, which would also increase.
More load shedding
Even with extensive use of peaking plants, there will be times when there will not be enough electricity, particularly in winter. There is load shedding in both scenarios, but with one of the two Koeberg reactors down for the whole year in 2022 for refurbishment, there will be over 800 hours additional load shedding in that scenario. The additional cost to the economy as a result will be in the region of R33bn.
How about some optimism?
Unfortunately, this is already making very optimistic assumptions about the life extension refurbishment. Looking back at recent history, we see that Eskom has never completed a major project on time. The Eskom estimate is that each unit at Koeberg will need to be down for 165 days for that work to be done, although at the same time Eskom have also said there are significant challenges in completing the work in that time frame, including that the “Outage planned scope does not match the planned duration”. This is a convoluted way of saying “too much work, too little time”.
Mallinson’s model simply accepts without question the Eskom estimate of R20bn for the refurbishment and assumes the project will be completed on time. More realistic estimates would make the refurbishment even less economically sensible.
Eskom acknowledged seeing Mallinson’s model at a media briefing on 27 January 2022, and the COO Jan Oberholzer said: “The document definitely raises an interesting perspective, however it includes a number of unsubstantiated assumptions around costing.” He also complained that Eskom was not consulted on any of the values used, but for some reason did not provide correct figures, and did not even say which of the costs were disputed.
The CEO Andre de Ruyter also responded to the modelling document describing it as “intriguing” and “quite interesting” but went on to say “within a year we can secure the future life of Koeberg … for another 20 years, which means we’ve got a highly reliable 24/7 capacity available at the most competitive cost in the fleet.” He went on to say that he still believed it was “a no-brainer” to extend the life of Koeberg.
Mallinson responds to the response…
“I was simply trying to show that what appears to be a fairly open and shut economic case to extend the life of Koeberg, is not actually that clear. In fact, keeping Koeberg running until 2024 may be a better strategy than taking it offline for extended outages to refurbish it.”
Asked if he would be willing to accept the correct figures from Eskom and include them in his model, Mallinson replied: “Absolutely. I would very much welcome the opportunity to work with the official figures directly from Eskom, especially the assumptions Eskom have made when modelling the cost of solar, wind and storage as replacements for Koeberg.” Mallinson is also keen to compare the calculations of the cost of the life extension refurbishment with the Eskom modellers.
Asked if he will share all his costing assumptions and calculations, Mallinson supplied the following table and said he was happy to provide more details on request.
In a letter appealing to other governments, five past Prime Ministers of Japan wrote on 27 January 2022: “Expert opinions from within Japan and beyond, have made clear that nuclear power is neither safe, clean, nor economical.” From an economic perspective, Mallinson’s report and the comments from de Ruyter it seems imprudent to incur costs and implement the life extension plan before the results of this modelling and the important perspective they provide have been carefully explored.
Hopefully Eskom will respond to Mallinson’s report with a little more detail than accusing him of using “unsubstantiated” figures. Eskom could also take the bold step of being more transparent and releasing its own costing analysis. However, given the history of secrecy around everything relating to Koeberg this seems to be very unlikely.
Here is the full modelling report: