When Gordon Hughes gave evidence to the House of Commons Science and Technology COmmittee the other week, I was struck by the vehemence with which his evidence was denounced as "balderdash" by Grantham Institute policy wonk Robert Gross. Unfortunately, Dr Gross's explanation of precisely why he thought this eluded me, as his subsequent narrative seemed to me to be little more than handwaving.
Being an inquiring sort, I decided to delve a little bit into Dr Gross's positions on wind power and started to read some of his publications. While I haven't got to the bottom of the dispute with Prof Hughes over the effect of wind on carbon emissions, I did find some interesting bits and pieces about just how you reduce intermittency of wind power by installing lots of turbines - a "wind carpet" in the jargon.
This has been the subject of a dispute between Gross and energy consultant James Oswald. In 2008, Oswald published a paper in Energy Policy which looked at this question and concluded that although there was some smoothing of supply, the effect was not nearly enough:
Although the aggregate output of a distributed wind carpet in the United Kingdom is smoother than the output of individual wind farms and regions, the power delivered by such an aggregate wind fleet is highly volatile. For example, had 25GW of wind been installed, with full access to the grid, in January 2005, the residual demand on the supporting plant would have varied over the month between 5.5 and 56GW.
As readers here probably know, if you balance your load with gas, this will mean much ramping up and down of output, with a concurrent loss of efficiency, so this is pretty bad news. However, Oswald went further and analysed the scope for balancing demand with supply from a wider area, but again found that this was not a plausible solution:
Wind output in Britain can be very low at the moment of maximum annual UK demand (e.g. 2 February 2006); these are times of cold weather and little wind. Simultaneously, the wind output in neighbouring countries can also be very low and this suggests that intercontinental transmission grids to neighbouring countries will be difficult to justify.
As his figure 13 shows, the correlation between different European countries can be striking:
The paper was clearly a serious blow aimed at the wind industry and elicited a response from Gross and some of his colleagues, and Grantham's men launched a strong counterattack:
Oswald et al. contend that their findings ‘significantly undermine the case for connecting the UK transmission grid to neighbouring grids’. This is because they argue that there is little potential for geographical distance to smooth wind farm outputs, because weather fronts create common wind conditions across large areas. Indeed Oswald et al. describe very large correlations in wind farm output across a wide geographical area based upon Met Office wind speed data. This finding is at odds with other empirical work, which finds that correlations between power swings from operating wind farms drop off markedly as distance
increases (Holttinen, 2005; Holttinen and Pedersen, 2003). We note that the sites chosen run in a line from north to south that for some reason neglect important offshore sites in the south-east. Moreover, other recent studies use data from operating wind farms and extrapolate as appropriate rather than using raw Met Office data alone (Ilex and Strbac, 2002). One reason for this is that geographically ‘low resolution’ wind speed data are not always a good indicator of local wind conditions and wind output. Future analysis needs to consider why it is that Oswald et al. appear to have found data that run counter to the established
view that distance between wind farms reduces correlations in output.
Interesting, I thought - a direct contradiction. Oswald finds a strong correlation between wind speeds across different areas, while Gross, citing Holttinen apparently says these "fall away" quickly.
The Holttinen papers are readily available on the web. Holttinen 2004 is a study that tries to estimate reserve requirements for systems with different proportions of wind in the energy supply mix, looking chiefly at hourly swings in supply and demand. On the face of it then, it doesn't even speak to the same point that Oswald is making, namely that we can get periods where the wind is not blowing strongly anywhere in Western Europe. Even on its own terms, the paper doesn't seem to support Gross's case. It says in its abstract that "Large geographical spreading of wind power will reduce variability, increase predictability, and decrease the occasions with near-zero or peak output." However, nowhere is there an analysis of how wind speeds correlate with each other over different geographical distances. In the text, there is this though:
When wind power comes from geographically distributed wind farms, the total production never reaches the total installed capacity and is hardly ever totally calm. From the combined production in the Nordic countries, production over 50% of rated capacity is rare in summer and production over 75% is rare in winter. The lowest hourly production was 1.2% of capacity for the Nordic wind power production time series.
Again, this hardly seems to make Gross's point.
Holttinen and Pederson is even worse. Although Gross claims to be discussing empirical studies, in fact this paper is purely theoretical, being a simulation of how a thermal energy system would react to the addition of large amounts of wind power to the grid. And since the system studied is the West Denmark grid rather than a wider geographical area, it again completely fails to address the point Oswald makes.
Perhaps I am missing something here?