National consequences of not understanding uncertainty

Business Report has an interesting article where Eskom recommends shelving new infrastructure projects due to lack of electricity until at least 2013. A few important quotes form the piece:

Bongani Nqwababa, Eskom’s finance director, said yesterday that the parastatal had advised the government that it wanted South Africa marketed only from 2013 for both local and foreign projects[….]  Eskom had not formally advised the department not to approve big industrial projects until 2013.

Which is slightly odd in itself – Eskom offering informal advice around electricity supply?

Eskom’s initial planning was based on 2.3 percent annual growth in demand for electricity. But this was revised to 4 percent when the government set a 6 percent economic growth target. Last year electricity demand grew 4.9 percent. In the year to March, growth is expected to reach 4.3 percent.

Three things strike me here:

  1. These demand growth assumptions seem to assume fairly substantial improvements in energy efficiency, assuming that as people become wealthier they won’t install air-conditioning in their houses and purchase more appliances, that offices won’t use more high-tech equipment pulling down Kws at every opportunity. I wish we had a graph of electricity consumption vs GDP.
  2. A few years of economic growth above the plan meant rolling black-outs? This sounds unlikely. I can imagine that the buffer would have shrunk, but not to the extent that is has. Also, as the buffer was shrinking, wouldn’t that be the first signal to up production, given the long time to get projects on stream?
  3. And if it really is a combination of higher than expected  growth and bad luck with Koeberg faults, power line problems etc. then surely this implies that the buffer chosen was too small? It doesn’t take an actuary to know that uncertain, dependent time-series processes (such as demand for electricity) can run away quite quickly. The more variation and dependency, the greater  the risk that the series (demand) will breach some bound (maximum sustainable supply). I’m sure that Eskom has a few engineers floating around and this stuff should be right up their alley.

If we add in the growth in residential housing developments, and the industrial infrastructure projects underway and planned (e.g. Coega) then the deterministic demand projections look on the low side to start with, and take no account of variation in this rate. If anyone at Eskom is listening, I’ll happily work with them to build a simulation model to determine what the appropriate size of the buffer really is.

Update:
A few more bits of information from an article written by Anton Eberhard from the UCT Graduate Business School

Annual peak demand has grown on average by just over 3.6 % per annum since 2000. Current peak electricity demand is actually lower than that predicted in Eskom’s Integrated Strategic Electricity Plans that were prepared in 2001, 2003 and 2005. Electricity forecasts have been reasonably accurate.

Furthermore, as far back as 1998, the Energy Policy White Paper warned of supply shortages around 2007.The claim that electricity demand has grown faster than predicted – and that this accounts for current supply deficits – is not supported by the data.

[…]

In 2001 Eskom was actually prohibited by government from building new generation capacity. Despite this embargo on Eskom, no concrete steps were taken for alternatives.

Interesting, yes?