Eskom grid collapse concerns well overdone

Apr 26, 2023

According to a research report from Kela Securities, the risk of a complete grid collapse and a nationwide blackout lasting two weeks or more in South Africa is extremely low. The country's electricity frequency of 50Hz is crucial for operation, and deviations from this frequency can cause blackouts. However, South Africa has automated load shedding and alternative measures in place to manage frequency and prevent blackouts. In the event of a blackout, power plants can be restarted within minutes using hydro plants and gas turbines, and the majority of the country can be back online within hours. The main challenge lies in repairing distribution infrastructure and coordinating supply and demand to restore power to all regions. The current situation is attributed to political interference and insufficient plant maintenance, which can only be resolved through proper maintenance and short-term load shedding. Here's the full report below.

By Lesedi Kelatwang and Kudakwashe Kadungure of Kela Securities

After having conducted on-the-ground research as well as consulting several experienced engineers with technical knowledge of Eskom's power generation operations, we at Kela believe that the risk of a complete grid collapse, leading to a blackout that will last two weeks or more, is extremely low. In fact, we understand that should the grid collapse, electricity would be back up and running within hours at the majority of the country's large economic hubs.

For a nationwide blackout to occur in South Africa, the country is essentially not producing any electricity at all and or Eskom's highly skilled managers have forgotten how to maintain the grid frequency 50Hz. Given that peak demand this year in May was c.33GW, and Eskom implemented stage 6 to get through it, a nationwide blackout is akin to Eskom implementing a further 27 stages of loadshedding above stage 6. This is naturally on the basis that each stage of load shedding is 1Giga Watt (GW).

Below we unpack how, at a basic level, this could potentially happen. Accordingly, we will also outline the reasons why we think it will not happen at all. We will limit our discussion to a single common parameter through electrical components, that of Frequency.

Electricity in South Africa is generated at a frequency of 50Hz. Using a dynamo to power the lamp on a bicycle as an example, where it is assumed to be rotating at the same speed of a power station of 3000RPMs (revolutions per minutes), it would consequently be generating electricity at 50Hz. As 1Hz is defined as being equal to 1 cycle per second, the dynamo operating at 3000RPMs would be achieving 50 cycles per second.

For South Africa, and in fact almost all grids across the world, it is sacrosanct that frequency is always maintained at 50Hz. As South Africans, our electrical devices/equipment are designed to operate using electricity generated at this frequency. A material deviation from 50Hz essentially means that anything connected to the grid would not work, or for the lack of a better word, break.

Power plants, together with household appliances can accommodate a small variation of frequency changes without any reaction. However, this is limited to a 1% frequency deviation in most cases, or 0.5Hz. This is known as the dead band, where frequency changes can be accommodated without any change in plant output. Any changes to the frequency of more than 1% will cause the plant to produce more or less power to bring the frequency back. And if the frequency cannot be brought back to within the dead band it will switch itself off. This is a safety feature that is built into the actual design of the plant.

Read more: Reserve Bank preparing financial contingency plan in case of total power grid collapse

What would cause a powerplant or grid to struggle maintaining 50Hz?

This can happen if there is an internal breakdown within the plant or if there is a material imbalance between the generation and utilization (supply/demand) of electricity. The former is pretty straight forward to understand hence we turn our attention to the latter. Every gigawatt (GW) generated must have an associated consumer (load).

Loosely speaking, if our theoretical grid of 5 power plants, with each producing 1GW generate 5GW, then you need 5GW worth of demand. In the instance of demand dropping to 3GW while supply stays at 5GW, this would have the effect of increasing the frequency of the grid. As mentioned, 50Hz must be maintained no matter what. The power utility would respond by reducing the power generated at all 5 power plants to be equal to 3GW, or switching off 2 plants, in order to reduce supply by 2GW. Alternatively the utility could look for means to increase demand by selling more electricity to neighbouring countries via the Southern African Power Pool (SAPP).

Conversely, in the instance of demand exceeding supply, this would have the effect of reducing the frequency of the grid. The utility can respond by increasing supply or by reducing demand. The former is not immediately possible for South Africa, and given that SAPP neighbours have relatively small installed capacity for South Africa to import from, this leaves loadshedding as the only option we have, which has the effect of reducing demand.

Assuming that all our plants are in perfect condition, what could possibly cause a black out?

In instances where supply exceeds demand the utility would need to switch off or throttle excess supply or export more to the SAPP. If, for whatever reason, this is not achieved then we would have several power plants taking themselves off the grid due to over-frequency protection, which would subsequently lead to a black out.

Logically, this is highly improbable given that Eskom has the control to switch off supply. In instances where there is a spike in demand that outstrips supply, the only way a blackout happens is when Eskom somehow fails to shed demand accordingly. While this is equally highly improbable, this has actually materialized in certain African countries. The underlying cause was that they struggled to cut off demand in time and the plants tripped, switching themselves off due to under-frequency protection as they are designed to do.

The reason these incidents occurred is due to there being a high degree of manual intervention in those countries in order to implement loadshedding. South Africa's grid is much more sophisticated in that there is a high degree of automation. South Africa has automated load shedding that is embedded in the system, over and above the normal manual load shedding we are all accustomed to, that is based on frequency to manage frequency decays and power plant generation can also be controlled centrally to reduce load or increase it, purely based on frequency values.

Over and above this, there is a wide array of alternative measures in place. For instance, Eskom can switch off some of the large industrial consumers for a few minutes. This is already accounted for in their contracts and can happen automatically when the frequency drops off sharply and the automated switching off of plants, based purely on frequency, occurs. In those few minutes Eskom can then fire up their reserves, namely their open-cycle gas turbines (OCGT), the pumped storage schemes and/or implement loadshedding in order to prevent more plants from tripping and allow for the frequency to normalise.

Read more: Power Shift: Eskom faces threat from independent producers as sales decline

How long would it take to get the grid back up and running in the event of a blackout?

Eskom has different types of power plants with each taking different times to start up. The coal and nuclear power plants take the longest to start up. Coal usually takes a few hours to start up while nuclear sometimes needs daysto get up to full load. The open cycle gas turbines and hydro plants take minutes to load up and are the natural preference for black starting a country.

Given that starting from a black out would be done using a combination of hydro plants and gas turbines, it would literally take up to a few minutes to get the power plant back up and running to some of the key loads. The rest of the country can be online within hours as more plants and loads get connected.

The challenge lies in coordinating supply and demand at startup in order to maintain the 50Hz frequency of the grid. This requires a careful process of slowly switching on and loading the plants whilst linking them to demand. In the first few hours the majority of the large loads would be reconnected. Cities and other large hubs would be reconnected during the first 12 hours while the majority of meaningful loads happen within 24 hours. What then would cause a blackout to last a few weeks?

Noting an observed trend where the increase in the frequency of loadshedding has led to a likewise increase in the number of outages lasting longer than planned, the cause is largely due to unforeseen consequences on distribution infrastructure. The increase in breakdowns of transformers and substations means that more time is lost to Eskom dispatching technicians and doing repairs onsite. In a scenario where loadshedding happens at once across the entire country, which in essence is what a blackout effectively is, it is conceivable that a certain percentage of the country would still be without power even though Eskom is back online.

Technically, in electricity generation parlance, brown outs in different regions. Surely, from a priority perspective, when switching the lights back on, areas of high economic importance would be high up on the list. Put differently, the likes of Johannesburg, Cape Town or heavy industrial areas such as processing plants or mining (for safety reasons) would be back up in a few hours. It would look like normal load shedding to those people. Whereas the poor bloke in a very rural area might have to wait a few weeks until Eskom eventually gets to him.

The key thing to note here is that a grid collapse can happen anytime anywhere in the world. Just because we have gone from stage 1 to stage 2 and now there is talk of stage 16 loadshedding does not mean the grid is about to collapse. Preventing a grid collapse is a function of maintaining 50Hz through active management of supply and demand of electricity. This is true whether you have 5 or 100 powerplants. In fact, avoiding an uncontrolled self- correction by the power plants, which are designed to be self-preserving when there is this imbalance between supply and demand, is precisely why you need to have loadshedding.

Naturally, the more power plants a grid has the more stable it becomes and the better the risk of trips and other issues can be shared across more plants. South Africa's 25 or more power plants with 100 or so generation units is in a decent position in this regard.

Read more: Government consider burning heavy fuel oil to alleviate load-shedding

With the above said what would a hypothetical stage 20 loadshedding look like?

With significantly higher stages of loadshedding, we expect things to get exponentially worse. Eskom's definition implies that stage 33 loadshedding is indeed possible, as this is currently where peak demand is. Given that we are currently at stage 6, there are a further 27 stages of deteriorating factors to go. Like most electrical devices, South Africa's power grid is not built to withstand the frequent switching off and on of power.

This then intuitively implies that stage 20 would entail much longer periods, and more occurrences, of switching substations on and off. Instead of loadshedding being a few hours it might be a few days at a time. The loadshedding methodology would obviously come under heavy scrutiny as the individual living in Sandton would argue that they should be load shed less than the individual living in Soweto as they are more economically relevant. Well, that could become an issue towards elections as political priority outranks economic.

How did we get ourselves into this mess?

South Africa is in this situation largely because of unnecessary political interference. More specifically, calls by politicians for Eskom to keep the lights on no matter what severely undermined Eskom's plant maintenance program they had in place for addressing the issues back then. The other reason being the decision by government in the 1990s for Eskom not to be involved in building more electricity capacity.

We look at an example where our theoretical grid of 5GW faces peak demand of 7GW while 2 of these 1GW plants are offline, undergoing maintenance. If maintenance on those plants was ‘prioritised’, meaning ‘do only the most necessary of maintenance’ to bring it online, they would be capable of generating a reduced 300MW and 400MW, respectively. The interference would go as follows:

1. Because of insufficient generation capacity, the country is in a situation where stage 2 loadshedding is required

2. Due to two plants being off for maintenance, there is an additional 2 GW of demand that must be removed, which leads the initial stage 2 to stage 4.

3. However, due to political pressure, an instruction would be made to keep loadshedding at say stage 3 no matter what. So, the utility would then look at the plants undergoing maintenance and react as follows: If the two powerplants, currently capable of producing a respective 300MW and 400MW can somehow generate at least 500MW each, then we can ensure that loadshedding is capped at stage 3. With that said the manager at each powerplant would then be approached and told to "do what needs to be done to get them to at least 500MW". The manager would in turn respond by saying "well, if I put a band aid here and another one there that would get me to 500MW".

4. The ambition to cap loadshedding at stage 3 would be achieved. However, this is short-lived as unforeseen breakdowns happen soon after due to insufficient maintenance. The problem will eventually show its face again as the power plants break down even more. Post the latest breakdown, powerplant 1 would go from 500 MW to say 200MW generating capacity, worse than before.

5. The situation then repeats; Due to political pressure, the utility is now told to cap loadshedding at stage 4 no matter what.

It should be quite obvious that this is a losing strategy. The problem can only get worse over time. Like a car that needs to be serviced in order for it to operate sufficiently over the course of its useful shelf life, the same goes for power plants. As such, planned downtime is required. In a situation where you have 10 plants on the grid, you should probably be running 8 plants at most at any given time as 1 plant undergo maintenance, no matter the demand, while the other is provision should another need unplanned maintenance. This was the strategy behind Eskom's 80:10:10 strategy.

In the event that demand surges to 12GW, you should still keep supply at 8GW to ensure proper maintenance whilst building more power stations in the background. Given the knock-on effect of loadshedding on government revenues, and therefore funding for Eskom, cutting corners on maintenance is a slippery slope. To do this in order to satisfy political pressure is a gross policy misstep. Of late, this policy is called ‘prioritising maintenance’.

Read more: Eskom's ludicrous expenditure: R238,000 for a mop – de Ruyter

How do we get out of this mess?

To fix the problem we need to properly fix the plants. Doing so naturally implies severe loadshedding as several units would need to be taken offline so they can be fixed. Hence much higher levels of loadshedding might become a reality. A reality which would last a few months potentially. However after those months, extreme and unpredictable loadshedding would be a thing of the past. In fact, loadshedding might disappear in its entirety.

For example, if Eskom sent out a notice today, saying that: they are implementing stage 10 loadshedding from September 2023 to December 2023; that hospitals, police stations and schools will be exempt; they are doing this to eradicate loadshedding in its entirety from 2024 onwards, would the average South African revolt against this plan?

This would however need to be well planned. Plant maintenance needs a lot of spares, most of which are imported. To really make a dent on loadshedding for the long term, the emphasis is again on plants needing to go under full maintenance. While we are mindful that the economy also needs to function, like the politicians reacting to their own needs, we think that short term pain for long term gain is the required response. But the question is whether the balance sheet and political willingness are there.

This then leads to the next question.

What type of person should be CEO of Eskom?

There are two basic criteria. The individual needs to be able to 1) withstand and manage political pressure and 2) be a proven implementor. The last thing we need is for a permutation of this aggressive strategy of increased loadshedding to be implemented and come 2024 the plants have still not been fixed and we have the exact same problem.

All Eskom power plants breaking down, at the same time… really!?

Going back to our initial discussion around causes of the blackout, the other plausible but extremely unlikely cause would be all the power plants breaking down to zero generation capacity. If this were to happen, it would be completely unforgivable. How on earth do all your plants break down at the same time? The risks are glaringly obvious. The solutions are equally obvious. If you need to escalate loadshedding to stage 10 so you can fix the plants and avoid a complete collapse, then do so. If there is indeed direct sabotage, then why is this so difficult to deal with? During covid, the South African government had it in them to dictate highly inconvenient measures to its population that even saw people lose their lives for not adhering to some of these changes. Yet, the same government cannot deal with a handful of criminals that are destroying the entire economy. Put different, this risk is blatantly easy to manage.

Overall, we think the scenario for a nationwide blackout is highly unfathomable. It is a very unique situation where never-seen-before incompetence at Eskom somehow coordinates the simultaneous breakdown of all plants and loss of imports from the SAPP. Under this scenario, much of South Africa would be without power for years, if not decades, as the state that is incapable of maintaining the plants today would clearly struggle to effectively replace them when the nation's resources are in an extremely worsened state.

Finally, what is our take on Karpowership?

We find the concept of procuring 1.2GW at a cost of ZAR200bn over 20 years, which implies a rate of ZAR10bn a year, or present value of c.ZAR72bn if using 20 year ZAR bond yields, to be highly illogical given that 4.8GW Medupi is reported to cost an estimated ZAR80bn. While we appreciate the time it takes to install and operate new capacity, this difference in pricing is, at best, highly questionable. As Kela, we are pretty confident the average South African would agree that a few extra minutes of daily loadshedding is a cheap price to pay to save ZAR200bn as a country.

Why not use the same funds to conduct proper maintenance and build new permanent generation capacity? This would be infinitely better for the country as Karpowership is akin to leasing a generator for your house at costs higher than buying a solar photovoltaic system.

Read also:

Listen to the story of Cyril Ramaphosa's rise to presidential power, narrated by our very own Alec Hogg.

Narration by Alec Hogg