For some countries – cold, windy ones like England – wind power appears to offer the best opportunity for displacing GHG-emitting electricity generation. In most developed countries renewable electricity generation from hydro is “tapped out” – i.e., there is no opportunity for developing further hydroelectric power.
There’s a lot of confusion about wind power. Some of this we looked at briefly in earlier articles.
Nameplate & Actual
The nameplate is not what anyone (involved in the project) is expecting to get out of it.
So if you buy “10GW” of wind farms you aren’t expecting 10 GW x 8,760 (thanks to DeWitt Payne for updating me on how many hours there are in a year) = 87.6 TWh of annual electricity generation. Depending on the country, the location, the turbines and turbine height you will get an “average utilization”. In the UK that might be something like 30%, or even a little higher. So for 10GW of wind farms – everyone (involved in the project) is expecting something like 26 TWh of annual electricity generation (10 x 8760 x 0.3 / 1000).
We could say, on average the wind farm will produce 3GW of power. That’s just another way of writing 26 TWh annually. So 10GW of nameplate wind power does not need “10 GW of backup” or “7 GW of backup”. Does it need “3GW of backup”? Let’s look at capacity credit.
Just before we do, if you are new to renewables, whenever you see statements, press releases and discussions about “X MW of wind power being added” check whether it is nameplate power or actual expected power. Often it is secondarily described in terms of TWh or GWh – this is the actual energy expected over the year from the wind farm or project.
The capacity credit is the “credit” the operator gives you for providing “capacity” when it is in most demand. Operators have peaks and troughs in demand. There are lots of ways of looking at this, here is one example from Gross et al 2006, showing the time of day variation of demand for different seasons in the UK. We can see winter is the time of peak demand:
If you have a nuclear power station it probably runs 90% of the time. Some of the off-line time is planned outages for maintenance, upgrades, replacement of various items. Some of the off-line time is unplanned outages, where the grid operator gets 10 minutes notice that “sorry Sizewell B is going off line, can’t chat now, have a great day”, taking out over 1GW of capacity. So the capacity credit for nuclear reflects the availability and also the fact that the plant is “dispatchable” – apart from unplanned outages it will run when you want it to run.
The grid of each country (or region within a country) is a system. Because all of the generation within most of the UK is connected together, Sizewell B doesn’t need to be backed up with its own 1GW of coal-fired power stations. All you need is to have sufficient excess capacity to cope with peak demand given the likelihood of any given plant(s) going off line.
It’s a pool of resources to cope with:
- a varying level of demand, and
- a certain amount of outage from any given resource
Wind is “intermittent” (likewise for solar). So you can’t dispatch it when you need it. Everyone (involved in producing power, planning power, running the grid) knows this. Everyone (“”) knows that sometimes the wind turns off.
If you add lots of wind power – let’s say a realistic 3GW of wind, from 10GW of nameplate capacity – the capacity credit isn’t 90% of 3GW like you get for a nuclear power station. It is a lot smaller. This reflects the fact that at times of peak demand there might be no wind power (or almost no wind power). However, wind does have some capacity credit.
This is a statistical calculation – for the UK, the winter quarter is used to calculate capacity credit (because it is the time of maximum demand). The value depends on the wind penetration, that is, how much energy is expected from the wind from that period. For low penetrations of wind, say 500 MW, you get full capacity credit (capacity credit = 500MW). For higher penetrations it changes. Let’s say wind power provides 20% of total demand. Total demand averages about 40GW in the UK so wind power would be producing an average 8GW. For significant penetrations of wind power you get a low percentage of the output as capacity credit. The value is calculated from the geographical spread and statistical considerations, and it might be 10-20% of the expected wind power. Let’s say 8GW of output power (averaged over the year) gets 0.8GW – 1.6GW of “capacity credit”.
This means that when calculating how much aggregate supply is available windpower gets a tick in the box for 0.8GW – 1.6GW (depending on the calculation of credit). This is true even though there are times when the wind power is zero. How can it get capacity credit above zero when sometimes its power is zero? Because it is a statistical availability calculation. How can Sizewell B get a capacity credit when sometimes it has an unplanned outage? We can’t rely on it either.
The point is, hopefully it is clear, sorry for laboring it – when the wind is zero, Sizewell B and another 60GW of capacity are probably available. (If it’s not clear, please ask, I’m sure I can paint a picture with an appropriate graph or something).
Low Capacity Credit Doesn’t Mean Low Benefit – And What We Do About Low Capacity Credit
Let’s say the capacity credit for wind was zero, just for sake of argument. Even then, wind still has a benefit (it has a cost as well). Its benefit comes from the fact that the marginal cost of energy is zero (neglecting O&M costs). And the GHG emissions are zero from all the energy produced. It has displaced GHG-emitting electricity generation.
What we do about the low capacity credit is we add – or retain – GHG-emitting conventional backup. The grid operator, or the market (depends on the country in question), has the responsibility/motivation to provide backup. Running a conventional station less often, and keeping a station part running, but not at full load – these reduce efficiency.
Let’s say we produce 70 TWh of electricity from wind (20% of UK electricity requirement of 350 TWh). Wonderful. We have displaced 70 TWh of GHG emitting power. But we haven’t. We have kept some GHG emitting power stations “warmed up” or “operational at part load” and so we might have displaced 65 TWh or 60 TWh (or some value) of GHG emitting power stations because we ran the conventional generators less efficiently than before.
We will look at the numbers in a later article.
So wind has benefit even though it is not “dispatchable”, even though sometimes at peak demand it produces zero energy.
Statistics of Wind and Forecast Time Horizons
Let’s suppose that even though wind is not “dispatchable” we had a perfect forecast of wind speeds around the region for the next 12 months. This would mean we could predict the power from the wind turbines for every hour of the day for the next 365 days.
In this imaginary case, power plant could be easily scheduled to be running at the right times to cover the lack of wind power. We could make sure that major plants did not have outages in the periods of prolonged low wind speeds. The efficiency of our “backup” generation would be almost as perfect as before wind power was introduced. So if we produced 70 TWh of wind energy we would displace just about 70 TWh of conventional GHG emitting generation. We would also probably need less excess capacity in the system because one area of uncertainty had been removed.
Of course we don’t have that. But at the same time, our forecast horizon is not zero.
The unexpected variability of wind changes with the time horizon we are concerned about. Let’s put it another way, if we are getting 1.5 GW from all of our wind farms right now, the chance of it dropping to 0 GW 10 minutes from now is very small. The chance of it being 0 GW 1 hour from now is quite small. But the chance of it being 0 GW in 4 hours might be quite a bit higher.
I hope readers are impressed with the definitive precision with which I nailed the actual probabilities there..
There are many dependencies – the location of the wind farms (the geographical spread), the actual country in question and the season and time of day under consideration.
We’ve all experienced the wind in a location dropping to nothing in an instant. But as you install more turbines over a wider area the output variance over a given time period reduces. A few graphs from Boyle (2007) should illuminate the subject.
Here is a comparison of 1 hour changes between a single wind farm and half of Denmark:
Here is a time-series simulation of a given 1000MW capacity in one location (single farm) vs that same capacity spread across the UK:
Here is an example from the actual output of the wind power network in Germany:
At some stage I will dig out some more recent actuals. The author of that chapter comments:
Care should be taken in drawing parallels, however, between experiences in Germany and Denmark and the situation elsewhere, such as in the UK. Wind conditions over the whole British electricity supply system should be assumed to be different unless proved otherwise. Differences in latitude and longitude, the presence of oceans, as well as the area covered by the wind power generation industry make comparisons difficult. The British wind industry, for example, has a longer north–south footprint than in Denmark, while in Germany the wind farms have a strong east–west configuration.
Here is an example from Gross et al (2006) of variations across 1, 2 and 4 hours:
Here’s another breakdown of how the UK wind output varies, this time as a probability distribution:
In another paper on the UK, Strbac et al 2007:
Standard deviations of the change in wind output over 0.5hr and 4hr time horizons were found to be 1.4% and 9.3% of the total installed wind capacity, respectively. If, for example, the installed capacity of wind generation is 10 GW (given likely locations of wind generation), standard deviations of the change in wind generation outputs were estimated to be 140 MW and 930 MW over the 0.5-h and 4-h time horizons, respectively.
What this means for a grid operator is that predictability changes with the time horizon. This matters because their job is to match supply and demand and if the wind is going to be high, less conventional stations are needed to be “warmed up”. If the wind is going to be low, more conventional stations are needed. But if we didn’t know anything in advance – that is, if we could get anything between 0GW to 10GW with just 30 minutes notice – it would present a much bigger problem.
Closing the Gate, Spinning Reserves and Frequency
The grid operator has to match supply and demand (see note for an extended extract on how this works).
Demand varies, but must be met – except for some (typically) larger industrial customers who have agreed contracts to turn off their plant under certain conditions, such as when demand is high.
The grid operator has a demand forecast based on things like “reviewing the past” and as a result enters into contracts for the hour ahead for supply. This is the case in the UK. Other countries have different rules and time periods, but the same principles apply: the grid operator “closes the gate”. To me this is not an intuitive term because he/she has contracts for flexible supply and reserves – in case demand is above what is expected or contracted plant goes offline. So gate closure means the contract position is fixed for the next time period.
However, the actual problem is to meet demand and to do this flexible plant is up and running and part loaded. Some load matching is done automatically. This happens via frequency. If you increase the load on the system the frequency starts to fall. Reserve plant increases its output automatically as the frequency falls (and the converse). This is how the very short term supply-demand matching takes place.
So the uncertainty about the wind output over the next hour is the key for the UK grid operator. It is a key factor in changing the cost of reserves as wind power penetration increases. If the gate closure was for the next 12 hours it should be clear that the cost to the grid operator of matching supply and demand would increase – given that the uncertainty about wind is higher the longer the time period in question.
Whether one hour or 12 hours gate closure makes a huge difference in overall cost of supply is likely a very complicated one, and not one I expect we can uncover easily, or at all. The market mechanism in the UK is around the 1 hour gate closure and so suppliers have all creating pricing models based on this.
Grid Stability – SNSP and Fault Ride-Through
System Non-Synchronous Penetration (SNSP) and fault ride-through capability are important for wind power. Basically wind power has different characteristics from existing conventional plant and has the potential to bring the grid down. We will look at the important question of what wind power does to the stability of the grid in a subsequent article.
Articles in this Series
Renewable Energy I – Introduction
Renewables II – Solar and Free Lunches – Solar power
Renewables III – US Grid Operators’ Opinions – The grid operators’ concerns
Renewables IV – Wind, Forecast Horizon & Backups – Some more detail about wind power – what do we do when the wind goes on vacation
Renewables V – Grid Stability As Wind Power Penetration Increases – Brief simplified discussion of Fault ride-through and System Non-Synchronous Penetration (SNSP)
Renewable Electricity and the Grid : The Challenge of Variability, Godfrey Boyle, Earthscan (2007) – textbook
The Costs and Impacts of Intermittency: An assessment of the evidence on the costs and impacts of intermittent generation on the British electricity network, Gross et al, UK Energy Research Centre (2006) – free research paper
Impact of wind generation on the operation and development of the UK electricity systems, Goran Strbac, Anser Shakoor, Mary Black, Danny Pudjianto & Thomas Bopp, Electric Power Systems Research (2007)
Extract from Gross et al 2006 explaining the UK balancing in a little detail – the whole document is free and well-worth spending the time to read:
The supply of electricity is unlike the supply of other goods. Electricity cannot be readily stored in large amounts and so the supply system relies on exact second-by-second matching of the power generation to the power consumption. Some demand falls into a special category and can be manipulated by being reduced or moved in time.
Most demand, and virtually all domestic demand, expects to be met at all times.
It is the supply that is adjusted to maintain the balance between supply and demand in a process known as system balancing.
There are several aspects of system balancing. In the UK system, contracts will be placed between suppliers and customers (with the electricity wholesalers buying for small customers on the basis of predicted demand) for selling half hour blocks of generation to matching blocks of consumption. These contracts can be long standing or spot contracts.
An hour ahead of time these contract positions must be notified to the system operator which in Great Britain is National Grid Electricity Transmission Limited. This hour-ahead point (some countries use as much as twenty-four hour ahead) is known as gate closure.
At gate closure the two-sided market of suppliers and consumers ceases. (National Grid becomes the only purchaser of generation capability after gate closure and its purpose in doing so is to ensure secure operation of the system.) What actually happens when the time comes to supply the contracted power will be somewhat different to the contracted positions declared at gate closure. Generators that over or under supply will be obliged to make good the difference at the end of the half hour period by selling or buying at the system sell price or system buy price. Similar rules apply to customers who under or over consume.
This is known as the balancing mechanism and the charges as balancing system charges. This resolves the contractual issues of being out-of- balance but not the technical problems.
If more power is consumed than generated then all of the generators (which are synchronised such that they all spin at the same speed) will begin to slow down. Similarly, if the generated power exceeds consumption then the speed will increase. The generator speeds are related to the system frequency. Although the system is described as operating at 50 Hz, in reality it operates in a narrow range of frequency centred on 50 Hz. It is National Grid’s responsibility to maintain this frequency using “primary response” plant (defined below). This plant will increase or decrease its power output so that supply follows demand and the frequency remains in its allowed band. The cost of running the primary response plant can be recovered from the balancing charges levied on those demand or supply customers who did not exactly meet their contracted positions. It is possible that a generator or load meets its contract position by consuming the right amount of energy over the half hour period but within that period its power varied about the correct average value. Thus the contract is satisfied but the technical issue of second-by-second system balancing remains..
..Operating reserve is generation capability that is put in place following gate closure to ensure that differences in generation and consumption can be corrected. The task falls first to primary response.
This is largely made up of generating plant that is able to run at much less than its rated power and is able to very quickly increase or decrease its power generation in response to changes in system frequency. Small differences between predicted and actual demand are presently the main factor that requires the provision of primary response. There can also be very large but infrequent factors that need primary response such as a fault at a large power station suddenly removing some generation or an unpredicted event on TV changing domestic consumption patterns.
The primary response plant will respond to these large events but will not then be in a position to respond to another event unless the secondary response plant comes in to deal with the first problem and allow the primary response plant to resume its normal condition of readiness. Primary response is a mixture of measures. Some generating plant can be configured to automatically respond to changes in frequency. In addition some loads naturally respond to frequency and other loads can be disconnected (shed) according to prior agreement with the customers concerned in response to frequency changes.
Secondary response is normally instructed in what actions to take by the system operator and will have been contracted ahead by the system operator. The secondary reserve might be formed of open-cycle gas-turbine power stations that can start and synchronise to the system in minutes. In the past in the UK and presently in other parts of the world, the term spinning reserve has been used to describe a generator that is spinning and ready at very short notice to contribute power to the system. Spinning reserve is one example of what in this report is called primary response. Primary response also includes the demand side actions noted in discussing system frequency..