Finding effective maintenance strategies for Eskom’s load shedding nightmare


A powerful video taken from Signal Hill in Cape Town last year showed just how serious the spectacle of load shedding is. As one-half of the city’s lights switched back on, the other half switched off as if a foreboding shadow had fallen over it. While the video was beautifully shot and quite satisfying to watch, it was a reminder that this is another grim reality that we as South Africans are having to bear.

As South Africa plunged into darkness once again in December, the blame seemed to be placed everywhere. Last year, it was the tropical cyclone Idai which damaged the transmission lines that carry power to South Africa from the Cahora-Bassa hydroelectric system in Mozambique. Then, it was the eight generator units that went down due to boiler tube leaks. Perhaps it was a combination of these and other factors, but the glaring issue and arguably the easiest to fix is the maintenance of power stations. The central problem, of course, has been that when Eskom is short of money, the power utility almost always tends to postpone maintenance because it simply is not a priority.

In an interview with Engineering News earlier this month, Eskom chief operating officer Jan Oberholzer revealed that the power utility had a new approach to maintaining its coal fleet, which will be considered by the board before the end of the month. The question then becomes, what exactly is Eskom’s game plan? In engineering, there are three main maintenance strategies deployed in the industry: run-to-failure, scheduled replacement and predictive maintenance.

Run-to-failure is where assets are used until they fail. For example, a conventional electric bulb is used until it burns out, or fails. Once the light bulb is out, there is a plan to fix it by getting a new one from either the store or from the cupboard and then replacing it at a convenient time. This strategy is often deliberate in that it is a cost-saving strategy but does not take into account lost production, customer unhappiness, re-work and other indirect costs in your analysis. You literally use the asset until it cannot be used anymore. 

This, of course, still requires a plan. Once the light bulb, for instance, runs to failure, there needs to be a plan in place regarding who should be responsible for the work, what parts they need and how they will carry out the task accurately and efficiently. In Eskom’s case, however, this leads to unplanned power cuts – precisely what we are trying to avoid. 

Run-to-failure is only a sound strategy when it will not cause a threat to life or have a significant impact on the company, and in this case, the economy. This, unfortunately, has mostly been Eskom’s strategy so far, which has resulted in running assets to failure without a plan in place for when this happens. The “emergency” created by unplanned power cuts requires procurement immediacy – this has potential for tender and other kind of infringements.

In scheduled replacement, assets are replaced when they reach the end of their life-cycle. This works for assets that cannot run to failure, such as an elevator cable. If an elevator cable runs to failure, it could result in serious injuries or deaths of the people using the elevator. In a scheduled replacement strategy, the elevator cable would be removed while it is still functional. The only real difference between scheduled replacement and run-to-failure maintenance is the knowledge about when replacement is going to happen. 

In this strategy, assets at Eskom would still work but would be timeously replaced to prevent unexpected power cuts. When there is a scheduled time for power cuts, it is because the power supplier is performing scheduled maintenance or replacement. Scheduled replacement, of course, is based on the theoretical rate of failure and does not account for actual equipment performance. This can often result in unnecessary maintenance. This seems to be the route Eskom might take, based on early indications from Oberholzer.

Finally, the maintenance strategy of the fourth industrial revolution (4IR) is predictive maintenance. In the electricity distribution networks, there is an infrastructure called a transformer. Transformers are classified into two types, and these are step-down and step-up transformers. For example, after electricity is generated and before it is put into the transmission line, it is transformed into high-voltage electricity by the step-up transformer. This is because it is much more efficient to move electricity over long distances using high-voltage. 

When this electricity reaches a town or city, a step-down transformer is used to convert high-voltage current to low voltage current. These transformers are costly. Transformer bushings are devices that are used to protect these devices. Unfortunately, bushings also are the source of the explosion due to a phenomenon called partial discharge.

Methods have been devised to predict when these devices are going to fail. An example of this is the work of a former Eskom engineer and Dr Sizwe Dhlamini (my doctoral graduate) who used artificial intelligence (AI) to predict the failure of transformers before they happen. Unfortunately, Dhlamini has emigrated to North America, a brain-drain trend that we need to understand and reverse. Dhlamini used AI and dissolved gas analysis to predict failures as well as the remaining life of transformers. In this way, one can use transformers that are otherwise scheduled to be replaced for a longer time, thereby reducing power outages and saving money.     

Predictive maintenance as implemented by Dhlamini is the route we need to increasingly adopt to create assets that are efficient from financial, technological and operational perspectives. We can generalise what Dhlamini proposed in his doctoral thesis to any electricity asset rather than just transformer bushings. In contrast to scheduled replacement, electricity infrastructures are directly monitored during normal operation to anticipate failure. Here, Eskom would be able to estimate when its electricity asset will fail and replace it beforehand. 

If you can predict when this will happen, you can get the most optimal use of them. The idea behind this is that there is convenient scheduling of corrective maintenance which helps to prevent unexpected equipment failures. This is more complex than run-to-failure and scheduled replacement. Eskom would be required to invest in condition monitoring sensors and other devices while employees will need to undergo training to use the equipment and accurately interpret the data they gather. However, this reduces costs and potential power cuts in the long run.

This is a compelling argument when you consider that the Council for Scientific and Industrial Research (CSIR), South Africa, has estimated that the cumulative cost of load shedding to the economy last year was between R59 billion and R118 billion. With the month almost up, Eskom is in a time crunch to announce its maintenance strategy. Failure to do so is another blow the economy cannot take.

For us to be able to achieve a more effective predictive maintenance strategy, we need to invest in several initiatives. Firstly, we need to invest in AI institutes that will create technology that can be used for predictive maintenance. These institutes should work with small businesses. Secondly, we should invest in data gathering and storage technologies. This will mean that the allocation of telecommunication spectrum should include companies that want to have their wireless networks to facilitate the gathering and transmission of data. Thirdly, we should invest in human capacity development. In the long run, these initiatives will place predictive maintenance at the centre of infrastructure maintenance across all sectors of the economy.

Professor Tshilidzi Marwala is a professor and the Vice-Chancellor of the University of Johannesburg. He deputises President Cyril Ramaphosa on the South African Presidential Commission on the Fourth Industrial Revolution.