Wind's impact on a power system can be assessed by examining demand swings from all consumers, first by pretending there is no wind in western Denmark and then by netting off the wind output. During this exercise, wind is treated as "negative demand." The largest power swings (750 MW and above) do occur more frequently with wind included, but the numbers are small -- 0.18% (16 hours a year) of the time with wind, and 0.05% in its absence (fig). Information of this kind is essential for estimating the additional spinning reserves needed for wind so they can cover expected fluctuations.
Operating utility networks is all about managing probability and risk. Greater probability of power swings means more uncertainty and the need for more reserves to cover the uncertainty. Margins for uncertainty do not, however, pile up on top of one another, Instead, the statistical laws of probability come into play.
A system operator managing a network the size of that run by the California ISO might have a forecast demand for one hour ahead of, say, 25,000 MW, plus or minus 300 MW. That is the "central estimate" of the error. It might be plus 600 MW, but with a lower probability. (All uncertainty margins include a range of estimates, each with its own probability). The operator will aim to schedule 25,000 MW of generation. If it has 10,000 MW of wind capacity on its system, generating 5000 MW, the best estimate of wind generation one hour ahead will also be 5000 MW. That leaves a need to schedule 20,000 MW of conventional generation. One of those units might trip during the hour and so that estimate also has error bands. The generation clearly will not be more, but might be deficient by, say 400 MW with a probability of 1%, or by 600 MW with a probability of 0.1%.
As for wind, data from Denmark, Germany and Britain all suggest the uncertainty margin on the estimate of generation (5000 MW) one hour ahead is plus or minus 300 MW -- the "standard error." The wind output might be minus 1700 MW, but with a very low probability. Based on these numbers and probability laws, the extra uncertainty due to wind, used to determine spinning reserve, does not call for 300 MW, but for about 60-80 MW.