A Few Thoughts on the Appropriate Risk Management Techniques for the Energy Markets
It is the Year 2020. Just one week ago Bob Carr Junior abolished the Electricity Tariff Equalisation Fund (ETEF). All technically developed nations (including Zimbabwe and Ukraine) have signed the Basel V International Accord establishing regulatory capital rules for energy producers and traders. Those who do not comply are not allowed to deliver energy to their customers or trade on the spot market and NEMMCO is forced to wipe them out of their dispatch software.
“First weigh the considerations, then take the risks”.
Field marshal Helmuth von Moltke
The next morning there is a blackout across the Australian East Coast …A sad but possible California-style scenario.
The recent Basel II Accord has established a relatively rigid framework of requirements for Risk Management for the banking and eventually, investment industry. Nothing even close to this Good Risk Management Practices yet exists for the Energy Sector.
The key factor in any risk management procedure is portfolio simulation and stress testing and there are several essential differences between the underlying Electricity Market and any other Financial Market that make the blind employment of the latter methodologies almost meaningless.
First of all the Electricity Market is driven physically rather than sentimentally. You may like or dislike Telstra shares, but you will never hesitate to put lights on when it is dark.
No matter what, supply should always meet demand, otherwise the physical grid will ‘protest’ and shut itself down. Only shareholders protested when trading of shares of OneTel was suspended.
When the demand in electricity is low, it is economically rational for the generator to continue supplying a minimum amount of energy and pay the penalty to NEMMCO rather than shutting ‘steamers’ down completely (plus starting and ramping them again with the rise in demand). This effectively creates a negative spot price, which turns most of the Log Normal models (nicely fitting the equity market) into useless mathematical exercises.
Certainly, numerous work-arounds have been developed in order to keep the familiar financial methodologies alive, but most of them turn into a miniskirt – by covering your knees you might expose something else… Ultimately, by employing familiar and widely accepted Financial Market methodologies to the Electricity Markets, you are increasing risk exposure instead of mitigating it.
Performing the mathematical modelling of electricity spot prices reminds me of a Russian saying: the further you go-the scarier it gets. A distinctly discrete set of half-hourly price readings with multiple spikes cannot be represented by a single-factor Geo-Brownian process. The combination of such a process with Poisson-like jump diffusion (and, finally, moving into a multifactor space) helps a lot, until you start fitting the parameters for your model and ask yourself a question: how do I distinguish in the discrete world a low-magnitude jump from the outliers of the Normal Distribution?
So, what would be the perfect framework? I believe, that in the absence of adverse events like plant outages (or another NSW Treasury Green Paper on Energy), there is only one independent stochastic driver – the system load. Even this is not purely stochastic, having a strong deterministic weather-correlated multi-cyclic component.
Therefore, to determine the future state of the market we need a deterministic Forward Curve together with the statistically defined Confidence Interval.
By incorporating the Forward Load Curve into the NEMMCO’s Linear Program Optimisation Algorithm (SCADA) one can resolve for the marginal generator and therefore, get the Market Price. All the necessary equations for this Algorithm are on NEMMCO’s website (just try to interpret them…).
This is, potentially, the ‘Unified Risk Management Theory’ (type of the Unified Field Theory which should explain every event in Nature by the same set of equations) that would answer all the questions. The first component of such a theory should be a reliable forecasting and simulating mechanism for the system load. The prerequisite for the latter is a reliable and ‘easy-to-handle’ statistical distribution.
As we could expect, the distribution of the raw load data (picture below) looks extremely ‘unfriendly’. It is a hunchback-type, fat-tailed, lumpy beast that would scare anyone who tries to approach it with the traditional analytical weaponry.
Therefore, somehow, the data should be transformed and massaged (or vice versa) providing at the end something symmetric, smooth and far more attractive.
Put simply, the ultimate goal is to obtain one single ‘nice’ analytical formula, with the minimum fitting parameters, which should not require speculations such as: this is a jump and this is just a giant step forward. Most likely, it must have fat tails, whose nature should inherently reflect the properties of the underlying Electricity Grid and indicate the dominant type of generation.
The only remaining question is: how, in the world, are we going to implement these requirements in a practical framework for risk management? We will tackle this question in part 2 of this article.
