There are lots of unfortunate side-effects of wind turbines on the environment, including effects on the electrical grid to which they are attached. It is no secret that the wind varies over time, and that variability translates to fluctuations in the electrical generation. These fluctuations occur on different time scales. Some years are windier than others, as are seasons, days, hours, minutes and even seconds. Over longer time scales these fluctuations are important because they necessitate having a reliable backup. Over shorter time scales they effect the reliability of the grid and increase the workload on the other generators – leading to a number of researchers to posit that this additional workload ends up increasing emissions to the point where there are no net savings at all. My personal perspective hasn’t changed in the years I’ve been doing this: we don’t know what the emissions savings from wind turbines are, potentially they are negligible, so why in the world are we spending billions of dollars when we don’t even know?
One problem has been getting information on just how variable this generation is. Ontario, for example, provides hourly numbers. For example, here is the output from Wolfe Island, a 198 MW project, for yesterday:As you can see, it varies quite a bit. Proponents would say that if you build a lot of projects “the wind always blows somewhere” and these fluctuations would “even out”. While this may be intuitive, it isn’t true. Here’s the same day for all of Ontario:In place of Ontario’s one-hour reports, some grids publish 15-minute reports. All of these numbers are handy when you are talking about market rules, i.e. lining up generation to meet overall demand. However, getting information on minute-scale or second-scale time periods has been quite difficult, and it is these fluctuations that make the other in-service generators work that much harder. To date, I’ve only seen one report that contained minute-by-minute numbers, in this case over a period of 3 hours:In Ontario this would be converted into 3 one-hour numbers, masking the wide underlying fluctuations. Sure, a “super grid” that could transfer large quantities of energy from wherever it is being generated to wherever it is being consumed is technically possible, but consider just how large and widespread (aka expensive and environmentally damaging) it would have to be. And even then it wouldn’t eliminate the need for reliable (fossil-fuel) backup.
In March 2013 Milan et al published a study titled “Turbulent Character of Wind Energy“. In it they studied the nature of second-by-second fluctuations and came to the conclusion that in order to successfully integrate wind energy into the grid you had to understand and mitigate these fluctuations. Unfortunately, there were no answers as to how this might be accomplished, and these authors cannot be dismissed as unqualified and/or anti-wind cranks. They are part of the Center for Wind Energy Research, Institute of Physics, University of Oldenburg, and this study was published in Physical Review Letters. Their goal is to find ways to successfully integrate real wind into a real grid, and Germany has had its share of problems with grid instability as they lead the world into this new technology. Problems that may not have practical solutions, which again raises the question of why we’re spending billions without knowing.
A rather good summary of the paper was published in phys.org and another on Oldenburg’s site. The graphic that appears in these summaries shows a tracing of the second-by-second fluctuations of output vs. wind speed for one turbine and it is quite remarkable. You’d think that given a certain wind speed you’d get a certain output, but this is not the case. Lower down in the phy.org summary is a video imbed that you’ve really got to watch. It runs for just 10 seconds and shows the generation over a 1o-minute period. I have a backup on youtube.
The study itself is just 4 pages plus a page of references. But it is a very dense 4 pages, as befits physicists writing for other physicists. If you understand, for just one example, what a multifractal cascade process is, have at it. But imbedded in all of the physics-speak are some interesting little layman-accessible tidbits. Below is what I gleaned out of the study.
My Layman’s Summary of Milan
Most traditional generators are rather large steam-driven turbines, where the electric current is generated by wires physically moving through magnetic fields. The frequency of the output (in North America, 60 hz) is determined by the speed of the turbine, and short-term (in the range of seconds) changes in load are partially counteracted and smoothed out by the inertia of the turbine. There exist very effective control mechanisms to control the speed and output of the turbine to satisfy current demand.
Wind turbines are designed very differently. Their main design point is not to match current demand; rather, it is to extract the most energy from whatever wind exists. For most modern wind turbines, that translates to an asynchronous generator, one that produces whatever it can with no regard for the condition or demands of the grid. This introduces the need for an inverter, where the AC of the generator gets converted into DC and then gets converted back to AC and synchronized to the grid. “In this decoupled configuration, the controller of the wind turbine commonly operates freely to maximize power output; i.e., to follow the wind power fluctuations mostly regardless of the grid load.”
When you’re looking at a wind turbine in operation the movement of the rotor looks quite constant, and one would think the generation is also constant. Over minutes the average output follows the average wind, but in the range of seconds it swings more dramatically than the wind speed, as demonstrated by the video. “Power increments look more intermittent than wind speed increments, exhibiting a clearer alternation between steady and gusty phases.” This is not your ordinary distribution curve, and a major part of this paper is to consider what mathematical techniques might be useful. “Power increments depart largely from the normal distribution, as they possess exponential-like heavy tails. These tails reach extreme values, corresponding to a higher-than-normal probability to record an extreme event. While wind speed increments look moderately intermittent, power increments are highly intermittent.” All of this is not good news if you are interested in maintaining a stable grid.
But wouldn’t this extreme variation be smoothed out by having more turbines? “The wind farm covers an area of about 4 km2, meaning that the 12 neighboring turbines are driven by similar winds, and they produce power outputs that are strongly correlated.” OK, so how about spreading projects over a wide area? “References [40,41] observe that such correlations are on the order of 600 km.” And “Reference  also observes power intermittency on a 300-km large wind cluster of 1000 MW at a time scale [of] 15 min , showing that intermittency persists at large spatial and temporal scales. This has a large impact on stability in largely wind-powered grids.”
Surely the geniuses that populate the wind energy industry have thought of all this and know how to fix it. Alas, “These aspects remain overseen in the wind energy community.” Are there any solutions? “Adapted solutions such as smart curtailment or energy storage remain to be developed with a focus on power stability. Such solutions must be properly dimensioned to match the extreme fluctuations observed. Otherwise, electrical intermittency and grid instability will grow larger with the ongoing exploitation of the wind resource.”
I don’t know about you, but solutions that “remain to be developed” don’t bring a warm glow to me, and they shouldn’t to anyone who has any grasp of the realities of our world. And I ask again, why are we spending billions on something of unknown value, especially when we don’t know how to make it of any value?
UPDATE – an Ontario perspective was offered by an engineer. It echos my understanding of the situation:
Sudden fluctuations (gusting) in wind power on a seconds to minutes timescale are taken care of by the kinetic energy of all the rotating turbine/generator masses on the grid, they would slow a little if the wind drops and inject energy into the grid to make up for the power loss. Grid frequency drops a little from 60 Hertz.
If a frequency deadband is exceeded normal turbine governor action kicks in to provide more output from the units on the grid, primary frequency control (except present nuclear units which operate turbine-following-reactor mode and do not contribute to primary frequency control).
If frequency offset exceeds a deadband Automatic Generation Control (AGC) will bring it back into spec. This is called secondary frequency control and is normally supplied as an IESO regulation service by hydro units at Niagara Falls although coal has been used in the past. It can also be done manually.
On a longer time scale the dispatching of stored water hydro and CCGTs at five minute intervals will bring the grid into the range that AGC can handle.
On a significant change in wind generation (intermittency) stored water hydro can be used to quickly balance the grid while CCGTs can be brought back into their dispatchable range. OCGTs are presently not needed on the Ontario grid to cater for wind, even more so now that the IESO will have the authority to dispatch wind.