Effect of a step change in wind output or load on in grid frequency and control mechanisms
The frequency is system-wide. That is, all parts of an A/C system will have the same frequency. Any change in frequency is a reflection of a change in the balance between supply and demand. So a drop in frequency is a reflection either of a loss of generation, or an increase in demand.
Actual demand at any time is a aggregation of many millions of demands, with all of us making changes, every time we operate a lift or switch a light or heater, or start or stop some process with motors, etc. So it varies stochastically, from the various changes we all make to demand. These changes can be considered random, but within reasonably defined probability distributions. This makes it feasible to make reasonable predictions of the short term average demand. But there will always be variations from this, and these fluctuations can be significant. Fortunately, we very rarely see synchronised actions that shift the demand by more than about 2% of the total. They do sometimes occur. There is at least a theoretical possibility of many of us all switching kettles on at the same time, and shifting the total demand by 10%. If we chose to set up a synchronising scheme so that millions of us all did something at the same time, the system would fail. But we can be confident that a very rare natural event.
If the imbalance remains, then the frequency will continue to change. So if demand stays more than generation, the frequency will continue to drop indefinitely. But changes in the frequency trigger other changes, both to supply and demand. Some of these changes are inherent, and others actively controlled.
A drop in frequency will generally reduce load. In part this is because motors will slow down, and so will consume less power, but there are other effects, perhaps including a general drop in average voltage, and so reducing resistive load. Nobody really know how much this happens, but there are guesses, and the effect is generally quite small.
A drop in frequency will also reduce generation, as the generators will slow down, and, without changes in torque (ie increased steam or gas pressure), will produce less power. This effect is considerably larger than the effect from load.
These two changes do not balance, so, without correction, the system frequency will continue to decline, and so collapse. The rate of decline is related to the inertia, mostly of generators, but also of load. It is also related to the proportion of generation that is lost, so a system with lots of small generators will decline more slowly than a system with a few large generators. Thus the size of the largest generator capable of instantaneous loss broadly dictates the “spinning reserve” that is needed.
It is this so called “spinning reserve” that has traditionally used governors to make the adjustments in output. When the frequency drops, the reserve generators are required to increase output. They should all do so by the same amount, so they usually share “droop”, commonly set to 4-8%. That is, when the frequency drops by say 1%, the increase in output should be 4-8%.
The governors should all react fast, in the UK within 10 seconds. In Ireland, which is a friskier system, as their generators tend to represent a larger proportion of the total, so it within about 5 seconds. The UCTE system can be slower. They are automatic, so have to respond to different events without knowing any particular cause.
These control actions, as well as covering for imbalance “events” also provide corrections to the stochastic changes in demand. So generators will be adjusting their output as frequency changes, albeit with a deadband, within which they do not react.
The System Operator will usually monitor the output of generators, so will be able to detect how much of the spinning reserve has been “used up”. If it gets a bit short (or long) they will then tend to replenish this by buying or selling in the Balancing Mechanism, so that generators will change their set points, and increase or reduce their output. These orders will not exactly match the changes in demand, so the changes they instruct can also cause a change in frequency, thus adding to the noise in the frequency signal, which is what you see.
So, to come to your question, how much electricity is represented by a given change in frequency, it all depends. Mostly it depends on the overall load on the system at the time. In a heavily loaded system, the loss of a generator will tend to be a smaller proportion of the whole than in a lightly loaded system, so a given frequency drop will correspond to a larger load. Broadly, in the UK, you can reckon that a 0.8Hz drop in frequency (down to 49.2Hz) will correspond to about 1 GW. In a lightly loaded system a 1GW loss might trigger a bigger change in frequency. But much depends on what the frequency response capacity and behaviour is that the time.
In general, spinning reserve is capable of delivering for only a short time: 10 minutes, or 30 minutes, depending. So it needs to be replaced as soon as possible, albeit by the “manual” action of instructing (or purchasing) STOR, at least until slower starting or ramping generators can be fired up.
Fridges are perfect for providing frequency response, as, at least for a while, they can collectively increase or decrease their load. But they do need replenishing, so that they can revert to normal when the STOR or other reserve is in place. This is something that is hard for system operators to get their heads around, as the fridges do this automatically, and not under their instruction!
One final complication about frequency. As I mentioned, averaged over a period of a second or so, the system frequency is a system wide signal. Wherever you are in the system, it will be the same. This does not mean the phase will be the same, so it may be that, when Scotland is exporting electricity to England, it will be ahead of England. Like a spring, the torque applied in Scotland will twist the spring a bit. Conversely, when England is exporting to Scotland, the spring will twist the other way, and England will be ahead of Scotland. When this changes, the spring will wind up or unwind a bit, and the instantaneous frequencies in England and Scotland will be slightly different, for a short while.
This introduces the possibility that generators, responding to the instantaneous frequency, will not all respond at precisely the same time. In the worst case, this can set up an oscillation, with generators in different regions swinging against each other. This makes the control of generators even more exotic and complicated.
Broadly, what this all means is that the control processes set up to handle stochastic variations in demand have the capacity to respond to short term stochastic variations in wind generation. Indeed, the capacity to cope with big frequency variations, which is essential for the safe operation (say) nuclear plant, will be an order of magnitude greater than that needed to cope with wind variation. So short term wind variation adds no new costs.
The bigger difficulty is with slower variations in wind, over hours and days. Broadly, 1 -2 hour forecasts fall with the error range of demand, so it is planning for 8 – 24 – 72 hours ahead that becomes more complicated.
But that is for another post.
!-!?!-Hirst Solutions Limited
Mobile: +44 7831 405443
From: firstname.lastname@example.org [mailto:email@example.com] On Sent: 20 September 2011 11:28
To: dave Cc: Claverton Grid; Claverton Wind energy group; Claverton AB MAIN GROUP; Kevin
Subject: Re: Effect of a step change in grid frequency