Deflating the numbers

The Imperial College (IC) paper wastes no time in bringing down to size exaggerated numbers presented by Hughes for the GWPF. As for the wind capacity needed to reach renewables targets for 2020, Hughes calculates it at 36 GW.

IC reduce that to 27 GW, on the basis of existing capacity (7 GW) and on- and offshore projects in the pipeline (20 GW), citing also estimates from National Grid ("gone green" scenario (pdf)), Department for Energy and Climate Change (roadmap for renewables), and the Committee on Climate Change (estimate pdf), none of which come close to 36 GW.

The importance of this becomes apparent when estimating capital costs. Hughes comes up with $124 bn for installing 36 GW (including the already built 7 GW!...) + transmission upgrades. On the basis of 27 GW, IC calculate overall capital costs as under £60 bn. Because of the 9 GW difference, but also because

Prof Hughes appears to have costed the entirety of his 36 GW at current offshore wind costs. His costing of transmission upgrading is not consistent with National Grid. For reasons discussed below, his estimate of `back-up' costs is questionable... (his estimate) appears to be some £64 billion above what the available evidence suggests, even without allowing for the possibility of cost reductions. (IC, p4)

Refuting the constraint fallacy

Prof. Hughes states that wind will "begin to impose increasingly heavy costs on system operation as the share of wind power in total system capacity approaches or exceeds the minimum level of demand during the year (base load). This threshold is due to be passed in the UK shortly after 2015."

Prof. Hughes makes a further contention that wind would need to be permanently constrained, to the effect that no more than 20 GW of 36 GW could be fed into the grid at any one time.

His calculations of wind farm economics are predicated on the basis that wind would therefore need to be substantially curtailed, which affects the load factor assumptions he uses.

Both comments show a fundamental misunderstanding of the statistics of wind output, and the means by which engineers assess power system balancing requirements. (IC, p4)

The IC paper argues that neither side of the equation is correct: the practical maximum of wind power is well below peak power, and the practical constraint on it is higher than the lowest demand during the year.

• Wind speeds are mostly to be found in the low to middle of the operating range, with the result that "wind farms most often operate at a range of outputs between 10% and 50% of peak output". Anything above 75% of nameplate capacity for the UK fleet of wind turbines is quite rare, thus even if such an event were to coincide with a demand minimum and wind would have to be spilt, it would not be a significant loss. (This point is based on research shown in Annex 3 of the IC paper, also covered at the end of the diary.)
• A key point is that there is more wind when demand is higher (the winter half of the year) and less when demand is lower (the summer half). As the IC paper says, the likelihood that it is extremely windy across the UK on a warm night in August (the lowest demand during the year) is close to zero.

In very simple terms, we can therefore assume that with 30 GW installed, wind power will produce, for much of the time, between 3 GW and 15 GW of power. Peak British demand exceeds 60 GW. Daytime demand in winter is typically about 45 GW, peaking to above 60 GW in the early evening. Minimum night time demand in winter is usually above 35 GW. In summer, daytime demand is usually around 35 GW and minimum night time demand about 25 GW. Wind speeds tend to be lower in summer, and higher in winter and autumn. Simply put, when wind data and demand data are looked at carefully we find that there is very seldom any need to spill wind. (IC, p5)

Further, simulations of the UK system run by the Department of Electrical Engineering at IC indicate little need to waste or "spill" wind (ie reduce turbine output compared to wind-speed potential; it's an old sailing-ship metaphor for taking wind from a sail because too strong or in an unwanted direction) up to 2020 and beyond. Note that all this considers the British grid only, without considering increased international balancing or energy storage (more on that later).

And now for back-up

After overestimating capital costs and problems caused to the system by (largely imaginary) excess output, Hughes got on to the familiar ground of the supposedly massive need for dedicated (and inefficient) back-up to take over when there is less wind output.

...wind needs 21 GW of dedicated gas fired back-up, all of which would be `open cycle' plant...

Open cycle gas turbines (OCGT) are peaker plants, easy to turn on and off. Their marginal costs are high because of fuel costs and low efficiency (compared to combined cycle gas turbine plants, CCGTs). "Whilst they are also cheaper to build, the capital cost savings are quickly overwhelmed by increased fuel costs, for anything other than the most short term uses." As a result, they are used for very short periods, generating small amounts of electricity when called on to meet a peak in demand or to back up in emergency situations. Current capacity in the UK stands at about 1.5 GW, they contributed less than 25 GWh to the British grid in 2011 - "the equivalent of running a mere 16 hours at full load over the whole year" (a load factor of just 0.18%!).

As the IC paper notes, it would be "economically absurd" to imagine them displacing the output of more efficient CCGTs, used for baseload or mid-merit generation. Systems operators are not going to call on OCGTs when they can get cheaper balancing power from more efficient plants. In other words, different plants produce according to their place in the merit order, and it makes no sense to consider wind power and a dedicated backup capacity separately from the rest of the power system.

Modelling by IC suggests that OCGT capacity needs to rise (to 10 or 12 GW) in response to occasional back-up needs with regard to increased wind but also phasing-out of some oil and coal plant used for peaking. Even so, they would generate electricity, as before, at a low capacity factor ("producing for an average of 30 full-load hours per year": a load factor of 0.34%) – in other words, they would continue to be used as peaker plants. The bulk of the balancing would be provided by CCGT plants and remaining coal-fired power plants: that is, mid-merit plants. (There is more detail, with power curves, in Annex 2 of the IC paper, which is also summarised near the end of the diary.)

This compares to the 21 GW, running for long periods, that Prof. Hughes cites in his evidence. We are not able to determine the basis on which Prof. Hughes calculates his estimate for the Committee.

The above outlandish scenario assumes that wind and its dedicated backup would replace existing baseload while all other generation would remain equal. If that weren't absurd enough, Prof. Hughes also has two even weirder scenarios of wind and its dedicated back-up replacing mid-merit resp. peak load (which we won't waste time on, but if you're interested, the IC paper takes them apart in its Annex 1).

It seems that in his supplementary evidence, Hughes revised the 21 GW figure to 13 GW. Perhaps plucking figures from a well-known bodily orifice doesn't always work. But the need, in terms of supporting data for a narrative, of a high capacity number and of higher marginal-cost OCGTs, is fairly clear when we see that Hughes pretends that a wind+gas scenario would be more costly, be less efficient, and emit more CO2, than a gas alone scenario. This kind of up-is-down narrative (You thought wind was low-carbon? Think again!) is often successful in impressing the media and public opinion, and this one has had a fair amount of play in Britain. Of course, it depends on specious assumptions, in particular that OCGTs would be turned on for long periods when in fact they would not.

Any other questions?

Looking beyond 2020, integrating a considerably larger share of wind power in the UK energy mix, as elsewhere, calls for adaptation of the system: transmission upgrades, storage solutions, smarter demand response. Hughes, for the GWPF, suggests that : "if the economics of such options were genuinely attractive, they would already be adopted on a much larger scale today".

The IC paper replies that, to some extent, they are – and points to grid connections between France, Germany, and Scandinavia that permit better integration of nuclear, coal and hydro. Also to time-of-day metering in France to help smooth inflexible nuclear output.

But, above all, why should one expect practices that were not previously called for to be already adopted?

Analysis by colleagues at Imperial College indicates that the potential for storage, transmission upgrade and demand response to reduce costs increases considerably as we look out to the longer term, to 2030 and beyond, and to a largely decarbonised power system. The BritNed interconnector to The Netherlands opened in April 2011, and the EirGrid East West Interconnector to the Republic of Ireland in September 2012. National Grid is proposing a line to Belgium that would be ready in 2018.

It may be that several decades of North Sea oil and gas windfall have restricted the British capacity for reasoning beyond the national sphere in matters of energy policy. Or at least, that a think tank that wants to put out a plausible but specious line can believe it may count on British insularity?

The merit order at work

Annex 2 of the IC paper discusses how the different modes of power generation contribute to the provision of variable power according to their merit order. This is illustrated by real and simulated power diagrams for Britain, all of which are reproduced below.

The first diagram shows the actual situation in the first week of November 2011, showing data for all plants directly connected to high-voltage transmission lines of Britain's National Grid. Wind doesn't appear significant because many smaller wind farms are connected to lower-voltage distribution networks. The production of OCGT plants is unnoticeable, while CCGT plants contribute to the provision of daily variation (mid-merit or intermediate load). The IC paper notes that this variability actually comes from the older, less efficient CCGT plants, so from the viewpoint of the merit order, it would make sense to separate CCGT plants into two categories.

The second diagram is based on a simulation with a few conservative assumptions: demand is scaled up for expected 2020 demand, wind power is the simulated total output of the 7 GW installed today (more on this in the summary of Annex 3), "other" (which included some international balancing) is ignored, and older and newer CCGT plants are now separate categories. The result: OCGT is still almost never used, instead, most of the extra balancing (above what older CCGT, coal and hydro plants provide) comes from throttling newer CCGT plants. But the newer CCGT plants still have a lot of balancing capacity left.

The third diagram shows the situation with the projected 30 GW of wind capacity. The time period includes an extreme case of high wind power during low demand, showing a small volume of over-production when wind would cut into nuclear's share. Meanwhile, there is less OCGT use than in the previous scenario:

The IC simulation model omits the consideration of some factors, most of which would further limit wind over-production and lower-efficiency gas plant production:

• import/export (although new transmission lines are being laid or projected, as quoted from the paper earlier);
• the capacity limits of transmission lines across Britain (but elsewhere in the paper they say there are projects for new domestic lines, too);
• the predictability of wind (which could be used to power down baseload plants);
• solar power (which would both shave off part of the daytime demand peaks and show some negative correlation with wind).

Wind load factor distribution

Annex 3 of the IC paper shows real and simulated load factor distributions for wind power in Britain, of which only the last is reproduced below.

The authors point out that data from the National Grid (which some anti-wind tracts have used recently) is only for wind farms connected to transmission lines managed nationally, and this data set is heavily skewed towards Scotland (where lower voltage National Grid lines provide direct connection for a higher fraction of wind farms). So they first simulated output for all of the 7 GW wind power capacity currently installed across Britain, based on real wind speeds. Then they went on to simulate the projected 30 GW capacity in 2020, with its higher share of off-shore.

The strongest change compared to the National Grid data (they used data for 2009-2011) is that low wind generation occurs much less often: the hours per year in the bottom two load factor bins (0–2.5% resp. 2.5–7.5%) are significantly reduced, with that for the lowest (0–2.5%) dropping from around 550 hours to around 150 hours. This result confirms the long-held contention that wind intermittency does indeed reduce with increased area.

It would be interesting if the Imperial College scientists were to redo the above analysis separately for different seasons, or for areas wider than Britain and the surrounding seas (whether just the entire Atlantic climate zone from Ireland to Poland or an expansion to the Mediterranean).