A key capability of any power generation and delivery system is to meet the peak demand for power. Fail to have either enough generating capacity available at those peak times, or the ability to connect and deliver this (via a grid) to customers, and their lights may go out. Technically, that’s known as “a very bad thing”…
One challenge is that building power infrastructure can become like building roads. As congestion at peak time clogs them up, we have tended to solve this by adding another lane to the highway (or more power stations and grid line capacity to the power system).
This is expensive. When peak hours subside and this expensive infrastructure sits largely unused, it also looks rather wasteful.
(Of course, for the companies who build this infrastructure and for the banks who charge to finance it, it’s an excellent way of doing things).
In a “smart” energy world there of course many solutions and services appearing which aim to address the impact of demand peaks, for example by moving electricity consumption away from the “rush hour” and to other times during the day. The use of storage, demand response services and time-of-use tariffs are just three such solutions – but these are topics for another day.
The first step in any approach has to be to understand the significance of the problem, by analysing the data: the drawing of “Load Duration Curves” is one useful method to do exactly this.
Example 1: the UK
Let’s start with some data for the UK, freely available from the National Grid here.
It gives data at half-hourly intervals (so 48 data points each day, reflecting the average demand in each period). Here’s a sample of what that data looks like, for the first twelve periods (i.e. midnight to 6am) of December 30th 2014:
If I plot this data in time order, now extending it for the whole of that day, this is the result: a typical daily “Demand Curve”:
I can do the same for the data for the whole of the last week of 2014 (with the 30th December indicated), to show how demand varied between different days. (I’ve removed the x-axis markings for ease of legibility).
And of course I can do the same for the whole year, plotting every data point in time order from 0:00 to 0:30 on January 1st through to 23:30 to 0:00 on December 31st. Unless I can stretch the x-axis out on some unfeasibly-wide screen, it produces something of a zig-zaggy mess! However it does show how demand in the UK is lower in summer than it is in winter.
A load duration curve takes this exact same data, but sorts it not into time order but into size of demand order. (Terminology alert: the term “load” – on the network – is being used here as synonymous with “demand” – from power consumers).
If we just took our previous data sample from midnight to 6am on December 30th, it isn’t dramatically different (since the demand does pretty much fall during that period):
Now let’s take the entire year’s data and do exactly the same: i.e. list every data point in order of size of demand rather than time. This will produce a load duration curve for the year:
This time I’ve left some numbers on the x-axis. They run from 1 to 17,520 and represent the ranking order of the data.
So the year’s highest demand during the year (in any half-hour period) is ranked “1”, with a value of 50,930 MW. The year’s lowest demand in any period during 2014 was 18,060 MW. So that’s last in the list, plotted on the right of the graph. Its rank is “17,520” because that’s how many half-hour periods there were in the year (i.e. a year is 8760 hours, so multiply this by 2).
So what we’ve done is simplify that zig-zaggy demand curve for the whole year by listing the data in size order, not time order. All the peak demand data points have moved over to the left and all the periods of minimum demand are on the right.
This allows us to much more simply draw some key conclusions around peak demand.
For example, the data point ranked 1001st in size has a value of 44,715 MW. Every point to the left of this has a higher value.
In other words there are 1,000 data points – rankings 1 to 1,001 – where demand was 44,715 MW or more. Now remember that each data point represents a half-hour time period. So this means that during the year there were 500 hours in total (1,000 half-hours) during which demand was 44,715 MW or above.
500 hours is 5.7% of the year.
The demand over these 500 hours ranges from 44,715 MW to peak (50,930 MW). This is a range of 6,215 MW, which is 12.2% of the demand peak (6,215 divided by 50,930).
In other words 12.2% of our capacity requirement is only needed for 5.7% of the time.
The shape of the load duration curve, steepening towards the left hand side, hints at the fact that if you consider fewer and fewer of the extreme demand periods, the disparity between capacity requirement and the time for which it is needed becomes more marked.
In the case of the UK data for 2014:
- The top 1,200MW of capacity was only needed for 50 hours, or 0.6% of the year.
- There were just 11.5 hours (0.13% of the year) with demand within 500MW of the peak.
Example 2: Steeper Load Duration Curves
The UK is not a particularly extreme example.
For example it’s been estimated that in parts of the US where air conditioning load is high, 10-20% of capacity requirements (generators & grid) is used for only 1-2% of the year; a few hundred hours (source)!
Worse still, peak generation usually calls upon the most expensively-fuelled power plants, since we leave them until last in the queue (“merit order”). As a result, peak power prices are not just higher than non-peak, but usually exaggeratedly so.
In Australia for example, where again peak demands coincide with the hottest days, the government reckons that “25% of retail electricity cost is accounted for by peak demand that occurs for less than 40 hours per year (less than 0.5% of the year)”; source.
By way of example, the next chart plots a year (2013) of load data for “profile area ACTEWAGL”* alongside the UK data, comparing the shape of the load curves. I’ve normalised the data such that the y-axis now shows demand as a percentage of the peak, rather than an absolute value.
(*“Actew Distribution Ltd and Jemena Networks (ACT)” is a network which covers Australian Capital Territory, around Canberra, plus part of New South Wales).
Compared to the UK, the Australian curve shows a dramatically steeper drop from peak values. Indeed, just as suggested in the US, the top 15% of load covers only 1% of the year and the top 20% of load just 2% of the year.
Interestingly, after a steep drop away from the peak the load duration curve adopts a similar gradient to that in the UK example.
With air-conditioning being one of the major drivers of demand growth in many developing markets, there exists a clear challenge to policy-makers in trying to ensure that they do not end up overseeing the building of large amounts of system capacity, only to see this expensive infrastructure lie idle for all but a few hours of the year.
Load duration curves provide one way to analyse the potential size of the problem and illustrate the need for smarter solutions, in particular those which keep peak demands under control.