"You often hear that the wind is too variable to contribute much to the electricity supply," says Arno Brand of ECN Wind Energy. "But when you take a closer look, it becomes obvious that the Dutch grid will easily manage."

Together with TU Delft, KEMA and others, ECN investigated the effect of large scale  wind farming on the balancing requirements of the Dutch electricity supply. The point of departure was an average scenario for the growth of the total installed generating capacity: slightly more than three percent per year, resulting in well over 35 gigawatt by the year 2020. In that scenario, Wind farms grow faster than any other power source. Their contribution increases to 7.8 gigawatt, 22 percent of the total capacity.
The wind itself cannot be controlled. Moreover, the low  marginal cost of wind energy - no need to buy fuel - makes it attractive to always maximize the power supplied by wind turbines. And since the Dutch grid has nearly no storage capacity, the balance between the active generators and consumers has to be closely controlled at all times. So when the wind fluctuates, generators elsewhere in the grid must increase or decrease their output to compensate.
It seems likely that new power stations will mostly burn natural gas rather than coal; a good thing, because gas-fired generators ramp up and down faster and more efficiently. Small scale generators like the many CHP installations in The Netherlands also help to maintain the balance, as does time-shifting part of the power demand. The Dutch transmission systems operator TenneT contracts in advance with industrial consumers for some of its power reserve.

The art of predicting
Power stations which can easily adjust their output are needed anyway, since power consumption also varies. The cycle of day and night, the seasons, changing weather conditions and unexpected events all help to ensure that TenneT never gets bored. And conventional electricity generation doesn't do precisely what's wanted either, without a little attention now and then. Maintenance is always going on somewhere, failures do occur. There is a certain variability in the system which is as unpredictable as the wind.
In other words, integrating wind turbines in the power system does not involve a fundamental change. It's just more of the same. What matters is the extent to which variations can't be predicted, because that determines the size of the required 'spinning reserve'.

Mathematical wind
Brand: "The question was, what is the reserve you need to live with a given amount of installed wind power capacity? And that makes two questions, because it concerns not only the difference between yesterday's prediction and today's real wind; we must also deal with variability - the random fluctuations around a correctly predicted mean.
To make calculations possible, a universal mathematical model was defined and supplied with local data. The KNMI provided wind records for sixteen locations; six offshore, four along the coast and six onshore. From these a full year of wind data was derived for the entire relevant area, and transformed to the hub height of the wind turbines.

Variations in wind power per 15 minutes, per hour and per 6 hours. The variation per 15 minutes is outside the scope of predictions. A reserve of ±14 percent is needed to compensate for the worst excursions within that time frame (illustration: ECN).

Space and time
"Our method deals correctly with the correlations in space and time," says Brand. "The model shows to what extend the variations in wind power cancel each other out or do the opposite, depending on the distances between wind turbines." The model also represents the wind farms as foreseen in the scenario for 2020.
At ECN we have large datasets of wind speeds and electrical power levels as actually generated by real wind turbines at our disposal, because we provide the daily wind power predictions for a number of wind farms. That is why Brand mainly concerned himself with the predictions within the model, taking also the structure of the balancing market into account.

A day in advance
In The Netherlands, the transmission systems operator (TenneT) works with "programme responsible parties" (PRPs) who buy energy for the providers, who in their turn supply electricity to their corporate and private customers. The PRPs are obliged to prepare daily e-programmes for TenneT, detailing the production and consumption of electrical energy from one quarter of an hour to the next, as planned for the following day. On the day of execution, differences between the e-programmes and reality will arise. The PRPs are each liable for their own programmes. If they can't match supply and demand, TenneT bills them; Electrical imbalances are measured in money.
"The model clearly shows the disadvantage of calculating per individual PRP," says Brand. "Each of them has one or at most a few wind farms in his portfolio of energy resources, which means that decreasing wind in one area is unlikely to be compensated by an increase elsewhere." The  required reserves were calculated twice. First by adding separate reserves for each PRP, next as a single reserve for the entire system, giving wind fluctuations more chance to cancel each other out.

Results
Wind turbines cause imbalance as the difference between the predicted and realized power levels. The required reserve equals that difference, en will be +58 and -56 percent of the installed power capacity when calculating per PRP. For the system as a whole it becomes +56 and -53 percent. So when 7.8 GW is installed in wind farms as expected by the scenario, the best case demands a reserve capable of 4.4 GW of up-regulation and 4.1 GW of down-regulation.
Imbalance is also caused by variability, short fluctuations around a mean. To balance those, a reserve of ± 16 percent of the installed wind power capacity is required when calculated per PRP. The system as a whole would need ± 14 percent.

Forward buying
"Those are large extra reserves," admits Brand. "But the study also shows how to make do with less. If the PRPs can still buy and sell energy on the day itself, then their wind forecast horizon will be much closer, reducing the error by half!"
Such "intra-day" trading now happens, and a cross-border intra-day market will shortly become operational. On June 9th this year the Amsterdam-based APX-ENDEX, Belpex and Nord Pool Spot agreed to create a market including Norway, Sweden, Finland, Estonia, Denmark, Germany, The Netherlands and Belgium. The infrastructure is also rapidly being expanded. A cable (NorNed) between the Eemshaven and Feda has been connecting the Dutch and Norwegian grids since 2008, the BritNed cable between the Rotterdam and Kent will come into use early in 2011, and there are plans for new connections with Denmark and Germany.

Economies of scale
"What you want is to extend both the market and the infrastructure from the Baltic to the Mediterranean," says Brand. "On that scale, forecast errors will also cancel each other out to some extend." An unexpected lack of wind here would then be compensated by an equally unpredicted excess elsewhere, requiring less up and down-regulation by conventional reserves. The method designed by Brand and his colleagues is very suitable for calculating such effects, when supplied with data for the entire area. "That would give us a quantitative handle on whether wind farms might then be able to contribute something to the base load," he muses. 

Contact
Arno Brand
ECN Wind Energy
Tel.: 022 456 47 75
E-mail: Arno Brand  

Text: Steven Bolt

Info
Variability and Predictability of Large-Scale Wind Energy in the Netherlands, an ECN report 

This ECN Newsletter article may be published without permission provided reference is made to the source: www.ecn.nl/nl/nieuws/newsletter-en/