[NB. This article is a summary of a 1-hour webinar I delivered. So if you have more time and/or prefer video format, you can find that here]
Peaks in electricity demand are problematic. We need enough electricity generating and delivery (grid) capacity to meet them. As we move towards low-carbon sources, these may not coincide neatly with times of peak demand. When the latter occur, we may therefore need to call on our most expensive sources of supply, which means peak power can be (very) expensive.
To illustrate that last point, consider this dispatch curve (sourced from here).
Fig. 1. A “dispatch curve” for electricity supply:
Because these curves usually rise steeply beyond certain levels of demand, it means relatively small increases in (already high) demand can introduce very big increases in the cost of the electricity supplied.
Peak demands are particularly problematic where the electricity system has to be designed to meet extreme events (and their extreme prices), even though these events happen rarely. It’s a very inefficient way to build infrastructure, requiring capacity to exist and be maintained, even though called on to operate on very few occasions. It’s no wonder that the owners of this infrastructure charge heavily when they get the chance! It’s also no wonder that many system operators and utilities are looking at whether supplying previously stored energy can prove more cost effective than firing up those most expensive generators when peak demands occur.
Electricity Demand in South Australia
To illustrate an extreme case, particularly around the frequency and duration of one market’s peak demands, I’m going to chart some data from South Australia, one of six regions overseen by AEMO, the Australian Energy Market Operator (the data I’m showing is sourced from their website, here).
It’s also a market where the importance of storage has already been recognised, Tesla and developer Neoen making headlines in 2017 by rapidly fulfilling a government tender to provide a 100 MW battery. Although the bulk of that battery was to address grid service and resilience issues, the battery has also been profiting from some of the pricing extremes that happen around peak demand.
Firstly, here’s a plot of demand and electricity market (“regional reference”/spot) price data for the whole of 2017 in the South Australia market, taken at half-hourly intervals.
Fig. 2. Power demand in South Australia, in 2017
The blue demand curve is noticeably variable, from a minimum of 560 MW up to maximum just above 3,000 MW. From April through October though, the variation in demand is noticeably more constrained and less volatile than it is during the summer months, November through March. During the latter, that blue line becomes very spiky, with distinct sharp peaks reaching above 2,500 MW (whereas March 2nd through November 28th – fully 272 days of the year – sees nothing above 2,290 MW).
The orange pricing curve is even more variable! Although there are a couple of extreme events in the middle of the year, the most common (and highest – up to an eye-watering ~$14,000 per MWh) are between January and March.
Here’s a different chart from that same data, focusing on demand and now picking out just the daily peaks. It clearly shows just how much higher the highest peaks are above both the annual average and the peaks that occur during the majority of days during the year.
Fig.3. Daily Power Demand Peaks in South Australia, in 2017
Another way to analyse demand data is to plot what’s known as a “load duration curve”. By plotting the demand data ranked in order of size rather than by date, we can better see how much cumulative time those spiky peak events take up during the year. The results may alarm you!
Fig. 4. Power demand in South Australia, in 2017, ranked in order of size by hourly period
If you haven’t seen a load duration curve before, then you may want to watch my video explanation here. Briefly though, there are 8,760 total hours in the year and the curve plots the total number of hours during the year that demand was above a certain amount. For example, the curve crosses the 2,000 hour mark at roughly 1,500 MW; meaning that there were 2,000 hours during the year during which demand was above 1,500 MW.
We are most interested in the left hand side, where the curve rises very steeply upwards. Indeed, from a maximum demand of just over 3,000 MW, the curve has already crossed below the 2,000 MW mark by about 250 hours. In other words, if we added up all the hours in the entire year in which demand was above 2,000 MW, then there are only about 250 of them (In fact, if we zoomed in closer, the exact amount is 282 hours).
That means that just 3.2% of the year accounts for the top third (33%) of our capacity requirement! Strewth!
Given the steepening of that curve as demand rises, it stands to reason that the higher the level of demand the rarer the occurrence. Sure enough, if we consider demand above 2,500 MW, that only happens during 37 hours in the entire year: one sixth of the system’s capacity is required for only 0.4% of the time!
Three days in February…
We saw from Figs. 2 and 3 that spikes in demand above 2,500 MW happened early in the year. Let’s narrow that down more exactly by plotting the exact days when demand rose above this amount, along with the number of hours on each of those days that this was the case.
Fig. 5. The top 37 hours of demand: how many hours above 2,500 MW on which dates?
I’ve also included on this same chart the peak price reached on each of those days (the yellow dots and the vertical axis on the right). Pricing is a result of a variety of supply factors, not just the level of demand: how much low marginal cost (renewable) power was available, for example. So you wouldn’t expect to see an exact correlation. Nevertheless, it is notable that, of the nine days shown, six of them saw prices peak around or well above $2,000 per MWh.
A run of three days, Feb 8th to 10th, account for more than half of the hours in the year which saw demand rise above 2,500 MW. I’ve highlighted them in red in the previous chart. They stand out on here too, a plot of demand and price data for the first two weeks of that month:
Fig. 6. The first two weeks of February 2017 in South Australia
Look back at weather data for that period, and it’s not hard to see why those peaks occurred. Temperatures in Adelaide on the 8th, 9th and 10th of February reached 42.4, 41.0 and 40.0 degrees Celsius respectively – those air conditioners would have been on full blast! From the 1st to the 7th, peak temperatures were lower, between 19 and 32; and the same was true for the 11th to 14th (maxing between 23 and 34 degrees Celsius).
Remember that demand above 2,500 MW accounts for less than one half of one percent of the time during the year, yet capacity of 500 MW – one-sixth, or almost 17% of the total – is required to meet demands between that level and the annual peak of just over 3,000 MW (which occurred in 2017 on February 8th).
With storage in mind, let’s ask this question: what would be required, on the three days above, to shave off that 500 MW capacity requirement and keep demand below 2,500 MW?
Here’s a repeat of Fig. 6 but showing just the portion of the demand curve above 2,500 MW on those three days:
Fig. 7. Peak power demands on February 8th to 10th 2017:
To shave those peaks (reducing our supply needs), what we want to know is two things. Firstly how much power we need to shave (the peak minus 2,500 MW) and secondly how much energy generation needs to be avoided (MWh: the area under each of those curves).
The results are shown below:
Table 1. Power and energy requirements to “peak shave” above 2,500 MW on three days in 2017
You’ll see I’ve included an additional calculation; “duration”, which is simply the ratio of energy to power. In other words, if we wanted to use battery storage – discharging it to replace peaking generators to meet these peak demands – we can see that an energy:power ratio of around 4 seems to be the requirement in each case.
It’s interesting to note that this number is one that comes up quite regularly in the storage world. And not just in Australia, but in other places too: anywhere from this gas peaker replacement in southern California to the UK National Grid’s capacity market auctions, where 4hr duration batteries are required in order to be awarded the maximum capacity rating.
In South Australia itself, although the Tesla “big battery” is usually described by its overall system size (100 MW, 129 MWh), it’s important to note that this is a partitioned battery, with 70MW reserved for short-duration grid services. The other 30MW is believed to have a duration of yes, you’ve guessed it, 3 to 4 hours.
Key messages & further notes
I’ve used quite an extreme example here to quantify some numbers around peak demand and illustrated in particular how significant supply capacity is required for really short periods of time during the year. South Australia is just one example market, but the same principles of analysis apply elsewhere. Whatever market you are involved in, it’s going to be crucial to crunch the numbers and quantify when peak power demands vary and on what scale.
It’s also important to think about both the opportunities and risks that apply to any business case that aims to “solve” peak power problems.
Consider storage, which is very much in vogue at the moment.
On the one hand, those high peaks and big price swings in South Australia look like attractive opportunities to sell previously stored energy at a premium price, as well as cutting out the need to maintain significant, underutilised generating capacity. On the other hand, the rarity of those extreme events would mean that, in the absence of other revenue streams to “stack”, a battery asset would end up rarely utilised or reliant on much smaller price-taking opportunities. We’d want to look at more regular day-to-day price swings and think about market saturation or cannibalisation too: how much storage could it take before demand (and prices) would be smoothed out to an extent that killed the very business case on which it had been built? Look at markets where storage is starting to eat into peak power supply – notably the US – and you’ll generally see contracts which de-risk the investment by being based on providing firm capacity (MW), not on energy price arbitrage.
Plus, of course, storage isn’t the only way to meet or manage peak demands! Flexible generators, demand-side response (DSR) and tariff incentives may prove to be more economic ways to address the problem. In practice most markets will see some mix of all of these, plus storage (in its various possible locations, from centralised to behind-the-meter).
I’ve only scratched the surface here. Nevertheless I hope that by looking at some actual numbers from one example market, I’ve given some initial insight into the kind of analysis approach that can help understand the drivers, dynamics and opportunity sizing that can be applied when investigating potential solutions to peak power in other markets too.