OzEA -- Electricity Market Analysis, Price, Demand and Value

Status: [27th Jan 2012] updated with 2010/11 data, and some general development of text.

Electricity Market Analysis, Price, Demand and Value

ABSTRACT and INTRODUCTION

Here is a high-level examination of the NEM market. In each NEM state wholesale electricity is priced every half hour by this market, and so we can take the price and the demand, in half hours blocks, and multiply: price x demand = value, summing up as required.

In what follows (so far) we look first at the overall value of the wholesale electricity market, and then at the concentration of that value into particular times.

DATA and METHODS

ANALYSIS and RESULTS

Working in billions (1e9) of dollars, we multiply out the value = price x demand of the wholesale market (in the half hour pricing chunks), and sum up for each region a year at a time. This calculation includes the occasions when the price goes negative.

So, the NEM wholesale electricity market is a seven to fifteen billion dollar a year affair. And we see that the blunt-average wholesale price of electricity is now sitting around $40 MWh. Note that $40 (per) MWh in the wholesale market contributes 4 c/KWh to the retail price (~20c).

Staying with the wholesale market, let's tease out the high-price parts, as it is these that contribute most to the average price, and it is around these that issues and opportunities arise.

The value in the extremes

The wholesale electricity price is capped at $12,500 MWh (previously 10K), which is around 300 times the average, and ~500x the median. Also, it can be at high-load times that the price rises to the cap; and a single half-hour can contribute more than 1% of the overall annual market value.

The following plots show, on log-log scaling, what percentage of the market value occurs in what percentage of the year:
(best if browser window wide enough to show plots in pairs)

As a choice of 'y-axis' threshold, we choose 1% (i.e. where the above curves cross the 10⁰ line) as this is safely above the region where many of the curves 'kick back', and thus provides a more robust measure than 0.1% (say) would.

So, looking at what portion of the market value occurs in the most valuable 1% of the year, gives the following:

It is striking that, for example, in SA in 2009, some 64% of the market value of 1.2 billion dollars occurred in half hour blocks constituting only 1% of the year. Overall, we see around 25% of market value existing within 1% of the time (as a blunt average view).

Note that for many of the curves above, reducing the threshold to the most valuable 0.1% of the year (~9 hr) does not make a big change to the contained portion of market value (i.e. for the curves that drop steeply to around or below the 0.1% mark before 'the kink').

REMARKS

1. There are other aspects that could usefully be examined, including the timing of the high value periods and their relation to demand etc. If and when someone gets to this analysis it will appear first in the comments.

2. In the above we look simply at what the data shows. Broader context is needed to understand 'why' the curves and numbers fall as they do. For example, the extremes of the SA market are presumably related to a combination of summer heat waves and a geographically concentrated population. Also, the general stability in Tasmania is presumably related to their hydro resources.

3. It is noteworthy that (from Table One) the total demand appears to have been growing modestly at around 3% from 2000 to 2005, and then we see a jump of around 10% to a level that grew slightly for a couple of years and now appears to be dropping. We can speculate that climatic effects (how hot it gets in summer; how cold it gets in winter) may be important here; we can speculate that efforts to improve energy efficiency may be succeeding. Any reader who can offer understanding or knowledgeable opinions on this is encouraged to comment.

4. The concentration of value in the extremes is striking. We see 25%, and sometimes much more, of the total value of the wholesale markets occurring in 1%, and sometimes much less, of the time. As pointed to in comment #2 below, the mechanistic basis lies in the bidding system through which generators 'line up' to supply electricity into the market. A sketch is given in Comment #4 for a simple model to explore the market dynamics, but thus far has not been pursued further. It remains that some open questions can be posed:
- What does the 'bend' / 'kink' in the above plots represent (seen especially in the 2009 and 2010 curves)?
- Is it healthy / common for a market to have so much of its value concentrated in this way?
- What incentives does this current market structure provide to market participants?
- What opportunities might exist for sharing in these extreme parts of the market? (including in particular demand management)

OzEA_ANEMVAL0002

Ben McMillan
Subject: Simple model for costs.
Date: 2010-12-13 (at 05:34:48)

One simple way of looking at this is to assume that there is a mix of generators with high capital cost and low fuel cost, plus generators with low capital cost and high fuel cost.

Most of the time, the electricity price is just the marginal cost of production (the low cost generators saturate first). Once the capacity limit is reached, though, the price goes to the cap, because any bid has to be accepted. This is when the generators recoup their capital cost.

In an efficient market, new generation should be built when the income made at the cap is enough to justify new plant, and this provides a constraint to find the total capacity economically justified. This explains why so much of the income needs to be generated in the small proportion of time when the system is near capacity.

Anyway, making some reasonable assumptions for the cost of plant and fuel cost gives you curves pretty similar to the ones above.

OzEA_ANEMVAL0004

francis
Subject: A very simplified model of different generators in an electricity market
Date: 2011-03-10 (at 16:08:49)

Some brainstorming following from comment #2 above;
we might proceed by considering some classes of generator:

1. Coal
2. Combined Cycle Gas Turbines (CCGT)
3. Open Cycle Gas Turbines (OCGT)
and
4. The Cap (abstract)

Taking the marginal cost as deriving from the fuel cost (per MW generated), which will be $12,500 for 'The Cap', and setting the fuel costs at reasonable values, it seems straightforward enough to explore the basic dynamics. Will also need to have values for the capital cost (per year) for each generator type.

We also take a demand trace for, say, a year at a time in a give market (a NEM state).

Let's start with 50% from coal, 30% from CCGT, and 15% from OCGT (leaving 5% at The Cap) of the maximum demand. The idea here is to iterate, modifying the balance of generators (i.e. starting at 50/30/15) in each step to achieve an optimum - this being where no generator is making a loss, and no generator is making a profit beyond that required to satisfy costs.

So, at each time point the generators bid to supply electricity according to their marginal costs. Any remaining supply needed after that available from coal and gas is costed at the cap price. As in the real market, all generators get the price of the highest offer accepted, and in this way the generators make money.

At the end of the year, after the marginal and capital costs, each generator has either made a profit or a loss. Now need to translate these profit/loss values into adjustments to the installed capacity... and a simple start here can be to translate the profit (or loss) into units of capacity (as per annual capital cost) and increase or decrease generating capacity by this amount. Iterate.

After some some number of iterations, this procedure will (hopefully) converge. With such machinery established, it will be interesting to do various parameter experiments. For example, to what extent does the value of the cap drive the division between generator types? How do changes in the individual cost parameters (the marginal and capital cost parameters for each generator type) effect the final result?

And then there are some interesting possibilities that arise with the addition of renewable generators.

Of course this is a very simple model, and for now I have simply been thinking it out a little. Might come back and implement in a few weeks, but that depends on where other thinking and priorities get to. Others are of course welcome to take off with this. Also, such work when established will need its own modelling page, but we can noodle about here for now.

Anyone got some reasonable numbers for the costs to hand?

OzEA_ANEMVAL0005

Ben
Subject: 5:30pm price dip?
Date: 2011-10-20 (at 16:01:59)

Hi all,

I've created an average diurnal spot price and demand graph using 11 years of data from AEMO web site (2000 - 2010).

The demand graph is fine, but the spot price graph has a huge dip between about 5pm and 6pm. Specifically, it drops from around $65 just before 5pm, to $40. It stays there for an hour, then shoots up to around $80!

I originally compiled the raw data into one file, cleaned it up, then used the longest excel formula I've ever seen to pull out every occurrence of each half hour time interval and average them. Thus I ended up with 48 figures, and graphed them.

I've checked as far as I can by eye if this dip is normal, but given there are over a hundred thousand cells of data it's difficult to tell. So, the question is have I stuffed up my formula? Or is this dip real? If so, why?!?!

Thanks. I notice no one has posted here for a long time, so hopefully someone is still checking here every now and then. Very useful site though!

Thanks
Ben

OzEA_ANEMVAL0006

francis
Subject: listening
Date: 2011-10-21 (at 02:30:34)

I'm here

OzEA news soon

happy to hear more about this analysis - lead us through the basics and the implications, and we can talk about the analysis offline (or here)

OzEA_ANEMVAL0008

francis
Subject: some perspective on the extremes
Date: 2012-01-29 (at 23:25:39)

To get some perspective on the extremes in the wholesale market (i.e. ~25% of value in < 1% of time) I've been looking at some artificial scenarios. The aim has been to get a rough handle on the 'value' inherent in demand management. This may become more important with higher levels of renewable penetration.

First, consider a participant in the wholesale market who has the capacity to 'switch off' and thus 'withdraw' from the market for short periods. What sort of saving can be made against the annual electricity bill by absenting from a small number of the high-value half hour blocks? The following plots address this question:

In constructing these plots I have:
- considered 1% of the year (88 hr) as the x-axis
- assumed a usage pattern that tracks overall demand, except when 'switched off'
- maintained the users total demand over the year

As such, I stress again that this is a very simplified view.

The regional and yearly dynamics observed in the head post are reflected in what we see here. As an overall 'yardstick' number, I say that a 20% reduction in an annual (wholesale) electricity bill could be achieved by 'switching off' for less than 20 hours a years (40 half hour blocks; ~0.2% of the time).

Presumably, many large electricity users already take steps in this direction.

For a retail consumer, suppose that the cost of electricity is the sum of the wholesale price plus around 15c per kWh. In essence we assume the consumer has access to the wholesale market, but still pays the retail overhead. In this case the plots (shown below) are qualitatively as above and with the saving reduced by a factor of ~4. That is, and as a blunt overall view, switching the house off at the fuse box for less than 20 hours a year could reduce your electricity bill by 5% (or more, or less, depending on the year and region).

Now; there is also value against the retail overhead in a consumer not stressing the network at peak periods, and I reckon it conservative to take this into account to achieve 10% overall as a ballpark figure. So, in my situation where I pay about a twelve hundred dollars a year in electricity bills, some demand management (if catered for) can save ~$100 a year, perhaps more. This need not be in the simple minded form outlined above. Such a pay-off could be worth investing up to a thousand dollars to achieve (in terms of domestic demand management infrastructure).

Perhaps all this gives some context to smart meters and demand management coming, but not coming especially quickly.

Finally, to reiterate, this is just an exploratory analysis with some 'yardstick' reckoning built over the top -- very happy to hear constructive comments and criticisms.

Name:*
Email:*
Website:
Subject:*
What is the longhand for 'Oz'? :*