Friday 17 April 2020

Negative Prices - a Rich Irony

I don't know how many people are aware of negative (wholesale) prices in the energy industry.  They often come as a surprise to the layman, but were easily and correctly predicted by those who understood the game, way back when electricity was first mooted as a candidate for being commoditised and de-monopolised into a traded market.  They would arise because some kind of pricing dynamics would be needed that disincentivised over-production.  

Why would this be an issue?  (a) The difference between peak demand (6 pm on a freezing winter's evening) and least demand (3 am on a hot summer's morning) is extreme, and both must be catered for, ideally via price signals; (b) quite a lot of electricity generation is "must run" - nuclear and run-of-river hydro being the traditional ones, but even back in the early nineties it was obvious that solar- and wind-power would be coming on in ever larger quantities.  (Of course, in those cases "must run" is by green-policy fiat, whereas nuclear is to a large extent a safety issue.**)  So there was always the logical possibility that must-run would exceed demand, at which point some generators whose marginal cost was zero might need to be offered a lower-than-zero price - i.e. paid not to generate.

And so it transpired, mostly in the German electricity market++ to start with.  I've written about German power dynamics before: they are bizarre (the Energiewende is absolutely inane) and the cogniscenti weren't even slightly surprised when it started happening there.  However, such was the widespread puzzlement in Germany itself^^, the leading technical stiftung in that country, Fraunhofer, started publishing minute-by-minute explanations of each and every instance.  (This was when the examples were fewer than one a month.  They soon became much more frequent, and the good Fraunhofervolk began to tire of doing these analyses.)   

Short periods of negative pricing are now utterly commonplace in electricity markets around the world.  (Dis)incentivising via price signals is not the only mechanism for balancing electricity grids, of course: grid operators typically retain the right to make generators switch off because they're told to.  However, in the case of 'green' electricity (and sometimes other generators) there is often compo to be paid - part of the less-visible subsidy regime, even if no premium is paid for the electricity they do generate.  But market mechanisms are best.

Rich Irony    Now we come to the funny bit.  The costs of constructing some types of renewable electricity generation sets are plummeting, especially for wind and solar; and of course the variable costs are close to zero.  In several markets, they don't need the traditional type of subsidy - a guaranteed offtake price - any more.  That's great.  (There are other, less obvious subsidies on offer, such as not being made to pay for the trouble their intermittency cause the grid.)

But here's the thing.  In many countries now there are auctions for who gets to build new generating plant, where bidding is for how much subsidy you'll accept - lowest bidders win.  Already, some applicants have bid in at zero, meaning they don't need a cash subsidy at all ...  Sometime soon, windfarms are going to start bidding in at "negative subsidies", i.e. they will be willing to pay in order to join the game!   

Ain't markets wonderful?

** Of course the very existence of nuclear power is by policy-fiat, too 
++ For afficianados, there was also an early (and quite unexpected) instance of a negative price in the UK gas market, where the underlying logic is the same, though conditions are rarely so extreme as in electricity.  It was, in essence, a freak occurence, albeit wholly explicable after the event
^^ As noted here before, even educated and professional Germans truly don't understand how markets work 

Technical Note (may safely be skipped):  a word on one of the technical spin-offs of all this.  Some readers may be aware that "traditional" modelling of the seemingly random and jerky behaviours of spot prices in commodity markets uses "geometric brownian motion" as the explanatory mathematical assumption to capture volatility, with its associated parabolic function.  It works brilliantly for post-rationalising & analysing price evolution across all the "traditional" commodities - metals, oil, agriculturals etc (in conjunction with a mean-reverting function, and sometimes a seasonality function and a drift factor).  It even worked for modelling natural gas when that started to be traded, notwithstanding spot prices in that market that were (and remain) vastly more volatile than ever before encountered.  But it doesn't work for power: (a) because the extremes of electricity prices include blips that are more even extreme than can be "explained" by a standard volatility function, however high you turn the dial; and (b) brownian motion can't "explain" / represent negative prices!  Another function was required.  I'm proud to say that Enron had people doing original mathematics: and to cut through the heavy stuff (though I can explain if you wish ...) the answer was an inverse hyperbolic sine (sinh-1), coupled with a Poisson distribution for the blips.

Now you know.


david morris said...

Thanks ND

Think I followed (most of) that.

Inane/insane ?

Nick Drew said...

Both, DM, I do assure you!

(but I intended what i wrote)

well done for keeping up!

old git Carlisle said...

I am ex grid controller (gas) and you've lost me.

Will read again !!!!

Nick Drew said...

You're not my old mate Ron Francis are you?

who gloried in the title "Central Controller" !

Anonymous said...

Oh go on then explain, you know you want to!

Timbo614 said...

ND: Thanks I think I got the gist. But not the technical note math!

How are these calculations/price forecasting methods coping with the changes and shifts in consumption patterns that are down to the virus?
Surely more power is being used in homes during the day and on the other hand huge industrial users have shut down. What would/might happen when 2 or 3 large cars factories switch back on? Or is that a distribution problem?

Nick Drew said...

Timbo - answer in 2 parts

Part (a): coping? - just fine: any reputable price-modelling approach recognises the possibility of paradigm-shifts (they happen every few years, sometimes big, sometimes minor) and of course we've just had one. All it means is, (i) identify when the shift occurs (sometimes easy because it's an obvious, identifiable event with a date-stamp on it; sometimes difficult at first to distinguish the start of the new paradigm-period from just being an outlier within the previous paradigm): (ii) obtain data from the new paradigm (or run with some hypotheses - maybe it started last week but maybe it was starting in a small way before Xmas): (iii) re-calibrate the model by tweaking the input variables until you get best fit (in these cases: strength of volatility & mean reversion; seasonality; maybe drift; variables to do with the blips): (iv) see whether "best fit" is satisfactory - don't be too quick to be satisfied. And keep repeating as more data become available - again, be critical of early efforts - until things seem to have "settled down" into whatever pattern has been identified.

Lastly, attempt to post-rationalise the new model - because au fond, there's an assumption that price patterns emerge for an underlying "fundamental" reason. Simple example: when we found that a Poisson distribution was brilliant for modelling the blips, we weren't remotely surprised because that's also how you best model (e.g.) the occurence of turbine failures! Coincidence? I think not.

Final comment on part (a): although many charlatans and feeble-minded "econometricians" cantilever out from this sound analytic framework to make forecasts, they are either pigshit ignorant or outright con-artists (and often both). The best they could ever say about such crap is: "provided nothing ever changes ..." which of course makes the whole "forecasting" exercise utterly, utterly empty.

(Their lawyers know better, and always add a little footnote in 6-point at the bottom of the forecast, as follows: anyone believing this crap to the point of doing something with it other than wiping their arses, needs to be certified as criminally insane)

Nick Drew said...

Part (b) - demand during lockdown: it's down overall, between 10-15%. Yes, the within-day pattern is different, but not nearly so much the geographical distribution (because most centres of shutting-down manufacturing & schools etc are close to where large populations live)

The Grid is pretty good at handling contingencies, certainly at handling reducing demand - and that's been happening steadily for a decade now; which means that, as regards the wires (high- and low-voltage) the whole UK system has quite a bit of spare capacity. We're broadly OK for generating capacity, too. So it's not an intimidating problem.

Specifically on factory restarts? In some cases for sure it's the Distribution system that bears the brunt of the changed pattern of consumption: but again, well within capacity. And what can look like quite a big electricity customer to the layman looks pretty small to the system. The ones that are individually noticeable even to the Grid (paper mills, metals plants, chemicals plants, glassworks etc) are always in detailed communication with the system, hour by hour.

The critical challenge in UK is mostly one of managing 6pm, which is a residential phenomenon. If the usual evening peak has gone up a bit (because more people are at home at 6pm / fewer in transit or down the boozer), well firstly the weather is quite mild, and secondly that peak of demand is most unlikely to have become higher than the kind of peak the Grid is well-accustomed to from events like when the World Cup Final is on telly. They have excellent data on demand from factories, schools, etc, so they will be busily calibrating what they are finding against what they already know. It won't be a problem, at least until next December and prob not even then.

It ain't gonna be energy shortages that do for us! (though EDF is likely to be in Big Trouble, for various reasons ...)

Timbo614 said...

Thanks for such a detailed and informative reply!
Co-incidentally: Coronavirus: Electricity use up during day as nation works from home

Which puts your analysis above bang on - you obviously love your subject.

It's like almost all disciplines from modern agriculture to high tech engineering. On the surface and to the layman or even experts in other fields they appear quite straightforward - until you scratch the surface and you are plunged into a morass of fine detail and complex issues.

Makes my coding life look pretty simple at times.

Anonymous said...

Im intrigued......why are EDF likely to be in big trouble?

Old Git Carlisle said...

No Ron has been dead for some time.

I was at East Greenwich with him and took over stage 1 HCRP some time after he moved to commission Stage 3 and thence move to Marble Arch . I used his steam superheater calculations during start ups on stage 1.

I moved fro Greenwich to Croydon and thence to Killingworth as Regional Grid controller.

I think I was the only Grid Controller not to incur the bulk supply tariff penalty. (more by luck than judgement!!)

Nick Drew said...

Interesting - thanks, OGC

It must be confessed that here on C@W I have sometimes quoted Ron, though (until now) not by name

during one of the periodic spats over BG's abuse of its monopoly (sorry to inflame you, but it's true, BG certainly did) Ron and I were having a beer and he said:

I don't understand: we are safe and we are reliable. What's the problem?

As far as I could tell, he genuinely didn't get it

still - lovely bloke

phil5 said...

"...the answer was an inverse hyperbolic sine (sinh-1), coupled with a Poisson distribution for the blips." Blimey, I don't remember seeing that in John Hull!