Wind power: not always there when you need it

The government's recent announcement on the carbon tax contained lots of information about how the government planned to encourage the use of renewable energy. However, the large amount of material produced for the announcement neglected some inconvenient truths about renewable, particularly wind energy system.

As is now becoming apparent, although wind energy may reduce carbon use in electricity grids (this point was previously unclear) they are strictly additions to the existing generators. Detailed figures from the wind farms spread over large areas are now available and preliminary analysis of the figures in both Australia and the UK show that no grid operator in their right mind would rely on wind for base load power.

The results clearly show that wind networks over large areas generate little or no energy for hours at a time. The networks may generate enough to take over from fossil fuel plants for varying periods, but grid operators will always have to plan to have enough capacity on hand to meet peak loads as if the wind farms did not exist at all. Spreading the wind towers over a large distance helps a little, but not enough to make a difference.

In other words, wind energy will not replace one bolt of one conventional power station, despite most attempts at costing this form of energy assuming that they will replace conventional plants….

One piece of evidence in this grim assessment is an analysis of the actual output of wind farms in the UK between November 2008 and December 2010 by Stuart Young Consulting. The report 'Analysis of UK Wind Power Generation' was paid for by a conservation body, the John Muir Trust. The trust's site doesn't say much about wind, but as its main focus is on preserving wild land and wild places, the report must be the result of concern about the effect of wind towers on the landscape.

The analysis found that on 124 separate occasions in the period, the output of wind farms taken onto the national grid fell below 20 megawatts – which is effectively nothing – from installed capacity of 1600 megawatts. Those nil power contributions lasted on average (note the word average) of just under five hours. Further, at each of the major power peaks during the year (such as half time during the FA cup, when everyone puts their kettle on), wind output was just a few per cent of installed capacity. In fact, although the average output of the various wind farms in the UK network was around 27 per cent for the study period, it was below 10 per cent capacity for more than one third of the time, and below 1.25 per cent for the equivalent of just under one day a month.

All this is different from more hopeful analyses conducted in previous years, but the report says "the nature of wind output has been obscured by reliance on 'average output' figures. Analysis of hard data from National Grid shows that wind behaves in a quite different manner from that suggested by study of average output derived from the Renewable Obligation Certificates (ROCs) record, or from wind speeds that are in themselves averaged." (The ROC reference is part of the UK method of tracking carbon reduction.) The report also notes that the existing UK pumped hydro capacity (hydroelectricity from dams) cannot possibly handle those blank periods.

Denmark, which uses wind extensively, routinely exports its excess wind generated electricity across the Baltic to Sweden and Norway, where it is used to pump water uphill into dams. The water can later be released as hydro electricity. The UK system does not have as many dams with hydro potential to hand, and Australia has very few.

The Australian electricity grids, and wind farms attached to it, are spread over a much larger area, so surely that can work to the advantage of wind farms? The problem is electricity can be transmitted only so far. A wind farm in Queensland cannot fill the gap if the wind in South Australia stops blowing. So then to have any real idea of the effectiveness of wind farms, as a first step we would need to know just how meteorology interacts with transmission distances.

Despite all the screaming and shouting about wind farms, very little work has been done on assessing this point. About the only piece of evidence, one way or another in this debate in Australia, of which I am aware, is a paper by weather analyst Andrew Miskelly and Tom Quirk, a former deputy chairman of VENCorp, which used to manage Victoria's transmission network. The paper, Wind Farming in South East Australia (Energy & Environment, 2010), looked at the generation figures for wind farms in South Australia and Victoria at five minute intervals for one month, June 2009. The figures are available for researchers from the website of the Australian Market Energy Operator (AMEO), which now operates the connected grids for Eastern and South Australia. The pair found that when the wind died, it did so right across the two states. Adding in wind from Tasmania helped a little, but not enough to make a difference.

The paper found that collective average output of the farms was about 30 per cent of installed capacity  (somewhat better than the UK experience) and, thanks to the spread of wind farms over two states, about 10 per cent of their capacity could be relied on for 90 per cent of the time. That's not bad for a bunch of wind farms, but it's still nowhere near good enough for a grid.

Electricity grids are run very conservatively and designed to cope with worst case scenarios. Grid managers cannot ignore the remaining 10 per cent of the time when the wind farms are producing nothing or very little, particularly if they do not know which 10 per cent. Nor can they brush aside the worst case involving wind generation – a hot, calm day – on the assurance that it does not happen all that often. They must plan for the worst, and have a generating reserve on top of that.

Then there are the question of sudden surges and sudden falls in output of wind farms. Miskelly and Quirk say little about those issues although the graphs in the paper show alarming changes in output for the wind farms. One response from the AEMO was to declare in 2007 that all new wind farms of more than 30 megawatt installed capacity would be classified as "semi-dispatchable". This means that rather than accepting the output from wind farms no matter what, the authority will set limits or caps on what it accepts over five minute periods. The caps will change depending on conditions (Semi-dispatch of significant intermittent generation – proposed market arrangements, AMEO, May 2007).

That would solve the problem of sudden surges in wind output destabilising the grid and greatly upsetting the many electrical appliances attached to it, but what happens when the wind dies away? Grid managers have to quickly replace any wind power that suddenly disappears from the grid or have a host of irate consumers on the phone. The common practice for all networks, long before wind energy came along, is to keep some "spinning" generators – that is, generators operating but perhaps at half power – all ready to be hooked up to the grid at a moment's notice, if other generators fall out.

The amount of spinning reserve kept in hand depends on a lot of factors including the risk that some part of the power on the grid may vanish, perhaps because a conventional plant develops a major fault, or perhaps because the wind dies. So how does wind energy affect the need for spinning reserve to be kept offline? Not only does this question affect costs, but the spinning reserves are generating carbon that the wind farms are meant to be offsetting. Almost the only real world evidence I have seen on this point, one way or another, is in the UK analysis sponsored by the John Muir Trust. The analysis in that paper states that the frequency of changes in output of 100 megawatts or more over a five minute period was "surprising". More work had to be done on the analysis but during just one month, March 20111 or immediately before the publication of the report, there were five instances of a five minute drop in output in excess of 100 megawatts, the highest being 148MW.

As noted 20 MW is effectively nothing but the average input from wind for that grid is around 500 MW (the wind input would only occasionally be anywhere near the maximum of 1,600 MW) so 100 MW represents about 20 per cent of the average wind output. Another way of looking at it is a small gas turbine plant has to be kept off line, in constant operation and so emitting carbon dioxide, to back up the wind farms. In some ways this is actually good news for wind, as previous estimates of problems by engineers had declared that virtually the whole output of wind would have to be backed up by spinning reserves. Instead, a more realistic figure might be about 20 per cent of wind farm output, with other gas plants ready to fire up at, say, 10 to 20 minutes notice to take over the rest of the output. At least that may be the case for the UK grid. The figures could be different in Australia because the grid is smaller, and wind may vary more or less, but this point has not been subject to public debate. The AEMO does not discuss this point in its published reports, and no one in authority has thought to ask it.

Among other efforts to deflect all this draining criticism of wind energy, activists point to wind forecasting systems. If the grid operators knows in advance just how much energy will be available, and when wind might decline, it can reduce the spinning reserves required. Considerable effort is being expended on wind forecasting systems around the world, but how successful are they? This is another in a long string of questions about wind that are not only unexplored, but mostly unasked. Very few of even the activists who support wind have no idea that such systems are being developed or why they are a good idea, let alone the general public or policy makers.

To fill in a small part of this information gap, a graph of forecasting accuracy over a range of time frames from the AEMO website is pasted at the end of the article. At first glance it seems impressive, but it leaves out a vital piece of information. How well does the forecasting system used compare with a simple status quo system – this is, simply guessing that in five minutes or an hour or whatever, wind speed will be the same as it is now? A status quo system would be extremely accurate until, of course, the wind changed. In any case, forecasting systems may help reduce the still substantial spinning reserve requirements for wind, but do nothing about the problem that as an energy source wind simply vanishes for hours at a time.

Activists may complain that it's unfair to produce two reports from groups that may have some bias against wind. Very well, where are the counter analyses by the wind lobby of the same figures showing that really wind is there all the time? In fact, where is any independent analysis of the emissions that wind energy systems might actually be expected to save in Australian conditions, versus the cost of those system? There is nothing but endless assurances, assumptions and hot air. The Australian government and public have jumped straight into wind energy with virtually no knowledge of how wind farms would interact with power grids, and without making any effort to find out.

Degree of error in wind forecasting systems. The lower the number the less error. This looks very good but is largely meaningless without some sort of comparison. Does this mean the spinning reserve requirements for wind have to be reduced, if so by how much? For that matter what were the reserve requirements in the first place?