Guest post by Alex Terrell and Andy Dawson:
Part 1 of the series on 2050 electricity demand provided a “high electrification” scenario where the average electricity demand was approximately 72GW, but peak demand on exceptionally cold days could reach 121GW.
Part 2 described how this demand could be fulfilled with a nuclear supply model. In Part 3 we have used the same demand model to show how this could be substantially fulfilled with wind and solar power; though relying on significant amounts of storage to match supply and demand, and gas (or biofuel) capacity to operate when storage is insufficient. A number of different scenarios are explored, with the preferred scenario laid out below, adjacent to the nuclear scenario for comparison.
The main source of energy is wind power, with a capacity of 280GW “land-wind” equivalent. If all of this wind power is offshore, due to higher capacity factors it would equate to 200GW.
Variations to the model have been trialled but do not improve on the solution or are infeasible:
- Adding realistic amounts of extra storage reduces the amount of gas that needs to be burnt, but makes no difference to the gas capacity required.
- Adding tidal flow and tidal lagoons in place of wind to the mix results in only a limited reduction in gas usage, and should only be done if it is cheaper than wind power.
- Converting surplus electricity to hydrogen and back again presents problems of storage and capital equipment.
- Converting surplus electricity to synthetic hydrocarbons could provide a route to a low carbon supply, but requires development of carbon capture technologies and significant capital expenditure.
There are far more variables to consider than in the nuclear model and therefore further permutations that have not yet been modelled. We have avoided a focus on costs because:
- They are highly speculative for deployment in the years 2030 and 2050.
- They would distract from the purpose of parts 2 and 3, which are to define the supply requirements for meeting the demand identified in part 1.
It is an easy task to add costs to the model. Agreeing on the costs though would be a more difficult challenge.
The demand target
Part 1 of this series modelled demand and identified the following electricity demand distribution curve for 2050.
Figure 1 – Frequency distribution of electricity demand in 2050. Whilst typical demand is around 60GW, extreme weather can push up demand (averaged over the course of the day) to 120GW.
This model already includes a reasonable amount of demand reduction under cold weather, high demand conditions. Therefore it is assumed that supply must meet this demand curve under all circumstances. It is acceptable – though economically damaging – to over produce and waste electricity. It is not acceptable to under produce, especially when people are relying on the electricity supply for their heating.
The Demand Model also assumed some heat is provided by district heating. In a nuclear scenario, we can assume that a small proportion of houses have the bulk of their heating provided by district housing. Wind and solar cannot provide such heating, so we have assumed that:
- There is a greater deployment of solar heating – including vacuum tube technology that is expensive, but can provide water heating in freezing conditions.
- The gas that is used is to some extent used in domestic scale fuel cells. These operate at a similar efficiency to the best CCGT plant, but at a domestic scale, their heat can be used more easily. Hence a large proportion of homes get some heat from these units.
The use of these technologies could be modelled further based on a variety of assumptions, but we have assumed they are neutral to the demand model.
Renewables data sets
For the renewables scenario, we have only considered wind and solar, with gas (or diesel or biofuels) back up that we aim to minimise. These are currently the only large scale renewables deployed for electricity generation (with the exception of biofuels). We have modelled the addition of some tidal based upon a mathematical model of tidal output over the month.
With the nuclear scenario, we considered 20 years of weather. However, with wind and solar, we have to match the weather with the historical capacity factors of wind and solar. National grid data sets record wind and solar output, and wind and solar capacity, so we can calculate a capacity factor. However, this is only reliable over the last few years of large scale deployment.
We have therefore used the National Grid records over the years 2012 – 2015 inclusive, to derive four years of wind and solar capacity factors on a half hourly basis. This data provides estimated figures for embedded wind and solar generation that does “not have Transmission System metering installed”. As such, it is national grid’s best estimate. Crucially, it also provides an estimate of installed capacity, allowing a capacity factor for wind and solar to be calculated on a half hourly basis, covering the geography of the United Kingdom.
We merged this data with the HadCRUT data for central England, and our demand model based upon the HadCRUT temperatures. We therefore have, for each day over a four year period, the computed demand, the solar capacity factor and the wind capacity factor.
Over these four years, the following minimum, maximum and average capacity factors were observed for wind and solar, and modelled for tidal and for demand:
Table 1 – Wind, Solar and tidal capacity factors 2012 – 2015. For the purposes of the model, demand and tidal capacity factors are assumed to be level over the course of the day – which will of course require storage and demand shifting to be covered later.
The average capacity factors shown are quite good – given that solar is generally assumed to give 10% in the UK, and Germany’s onshore wind fleet achieves a capacity factor of 16 – 18%.
The capacity factor of wind will increase as more and more wind is built offshore – current offshore farms achieved about 37% capacity factor in 2014, compared to 26.1% for the embedded (presumably all onshore) wind farms. It is not clear whether the distribution of capacity availability will improve – for that we would need to evaluate daily output from offshore wind only, and National Grid does not break this out.
Without evidence that output is less variable, then we can assume that future offshore wind is worth 37/26.1 (= 1.43) times as much as onshore wind. So, for example, 280GW of “onshore” capacity could be replaced by 200GW of offshore capacity.
A note on variability
We have often seen comments along the lines of: “Demand is variable anyway, so the variability of wind and solar is less of an issue”.
Clearly this is not the case. Coupling up a variable supply to a variable demand is more difficult than if one of those if steady. This is especially the case if the variability of the supply is greater than the variability of the demand. To measure the variability, we use the coefficient of variability, defined as the Standard Deviation divided by the Mean. The grid watch data for 2015, and our demand model for 2050, allow the following Coefficients of Variance to be calculated:
Table 2 – Coefficient of variation for demand, wind and solar outputs
Currently, demand has a fairly low Coefficient of Variance of 13%. In the 2050 “high electrification scenario”, it is increased to 19%, mainly due to the seasonal demand of heating. Wind and solar outputs however have Coefficients of Variance of 57% and 65% respectively. Combining wind and solar improves matters a little bit – the combined Coefficient of Variance falls to about 47%. However, the supply variability is still significantly more problematic than the demand variability.
Over production with renewables
To power the UK predominantly with renewables, it will be necessary to over produce and spill (or export) some of the surplus power. The following table shows the amount of production over a 4 year period – based on 2050 demand and 2012 to 2015 capacity factors – for given amounts of wind and solar capacity. The figures are as a percentage of demand, averaged over a four year period.
Table 3 – Percent of demand fulfilled by different levels of wind and solar over the course of a 4 year period.
For a base case, 280GW of wind capacity and 100GW of solar capacity would produce 121% of the UK’s electricity demand, over the four year period.
Wind and solar correlation
Wind capacity factors in the UK are higher in winter – and therefore negatively correlated with temperature. Solar is obviously positively correlated with temperature.
Figure 2 – Correlation of wind and solar capacity factors with temperatures
Wind power is actually over correlated with reduced temperatures – in that it will over produce in the winter and under produce in the summer. Whilst this is a good attribute for the UK, the random element of wind means this is not always the case, and in particular the correlation may not continue at very low temperatures.
Figure 3 – At very low temperatures, the correlation no longer holds, and is indeed (non-significantly) “positive” (which is not what we would like).
We have used data over the years 2012 to 2015, but as an anecdotal point, the capacity factors of wind and solar on Monday 20 December 2010 were 14.1% and 2.4% respectively. This was the coldest day over the last 20 years – with an average temperature in central England of -7 degrees C. If this temperature is experienced in 2050, then according to our model, demand for the day would be 2,905GWh.
Our base renewables supply model assumes 280GW capacity of wind power, and 100GW of solar capacity. On average, this produces 20% more than demand. However, based on the weather conditions of our coldest day, it would have produced 1,007GWh, a shortfall over 24 hours of 1,900GWh.
Base case: Wind, solar and gas
Any shortfall can be made up with gas, or an alternative storable fuel. This could be biofuels, if available in sufficient quantities. However, biofuel might be needed for aviation, shipping and long haulage freight, so, as with the nuclear capacity augmentation, we assume gas will be used. Storage can also help, as discussed later, but in the first instance we will assume no storage beyond that needed to level daily demand.
How much gas? Figures 4 and 5 show daily winter and summer supply applied to the 2050 demand model. This is based on calculations every 30 minutes, assuming that demand is levelled over the course of 24 hours.
Figure 4 – 2050 supply and demand based on winter 2014/2015 weather, with 280GW of wind capacity and 100GW of solar capacity. If there is a shortfall, it is made up with gas (shown in red). If there is a surplus, it is lost.
Figure 5 – 2050 supply and demand based on summer 2015 weather, with 280GW of gas capacity and 100GW of solar capacity. This is calculated on 30 minute intervals – within a day, there can be both excess of renewables and a shortfall made up with gas.
With varying amounts of wind and solar, the % of demand that is fulfilled by gas (or diesel, coal, or biofuels),and the capacity required based on 4 years of weather and capacity factor data, is shown in the following tables. These are based on 4 years of settlement period data, covering over 70,000 data records:
Table 4 – Percent of demand fulfilled by gas over a four year period, with storage.
Table 5 – Gas capacity required to ensure supply based on four years of weather patterns, assuming no storage and demand levelled during the day.
We see that in this instance almost the peak demand has to be covered by dispatchable capacity. Our base case characteristics are shown below:
Table 6 – Base case scenario
Our base case shows gas making up about 20% of supply, and 41% of supply being wasted (or exported). Clearly not a satisfactory end model!
Storage and intra-day variations
Adding a bit of storage to the calculation will improve matters quite quickly, as intra-day variations can be accommodated. However, storage provides diminishing levels of return.
The model caters for excess production to go into a storage reservoir, and this can then be returned. The assumed round trip efficiency is 75%, which is typical for pumped storage. That will also be fairly accurate for distributed battery storage, where there are losses in conversion from AC to DC to chemical energy, and back again. Based on this model, gas usage and required capacity were evaluated for different storage amounts:
Figure 6 – Gas usage and capacity versus different storage amounts
We need to stress that the horizontal axis represents vast amounts of storage. By comparison, Dinorwig has less than 10GWh. Thirty million households, each with a 30KWh battery, would provide 900GWh. In Sustainable Energy without the hot air, David MacKay asked whether Britain could store 1,200GWh with pumped storage – the answer, with a lot of difficulty:
“By building more pumped storage systems, it looks as if we could increase our maximum energy store from 30GWh to 100GWh or perhaps 400GWh. Achieving the full 1200GWh that we were hoping for looks tough, however”
What Figure 6 shows is that realistic levels of storage have no impact on the thermal capacity required. Based on the demand levels of Part 1 – and wind and solar output from 2012 to 2015, we will need about 104GW of thermal capacity in 2050.
We may actually need more than 104GW. When the storage is empty – which it will be at times – the full capacity is required. The 104GW is based on 4 years of capacity factors, but we need the supply infrastructure to cater for exceptional circumstances. In practice, that will probably require 115GW of available capacity. As in the nuclear scenario, some of this will rarely – if ever – be used, and could be provided by diesel or “mothballed” coal plants.
Although storage can’t alter the required backup capacity, it can improve the amount of fossil fuels burned. Figure 6 is recast below using realistic amounts of storage.
Figure 7 – Gas usage and capacity with realistic storage amounts. The red line indicates a potential inefficiency in running gas plant on a cycling basis.
Note also that we will need significant storage amounts – probably over 100GWh – to enable Combined Cycle Gas Turbines to run for extended periods at optimal efficiency. Otherwise, their efficiency goes down and emissions go up, as appears to be happening in Ireland . The effect of this inefficiency would be to increase the amount of gas consumed, as illustrated by the red line in Figure 7.
Intra-day demand management modelled as storage
The model above assumes that demand is levelised over the course of the day. As discussed in Part 2 (the nuclear scenario), with smart heating systems and a high thermal heat capacity in homes and offices, this is quite feasible.
We can use the same mechanism to shift demand for a renewables scenario. It is more complicated to implement, as heating will applied at – from a user perspective – random times of day, as opposed to regular times. We can model this time shifting as a virtual battery, with the following assumptions:
This is in effect allowing the temperature of homes to be lowered by 1C to store the equivalent of 450GWh of electricity. We can add in commercial buildings, subtract those who refuse to participate, add in those who are prepared for a 2C temperature change etc, etc. For the model, the result is storage – but only available when heating is applied.
Electric Vehicles for storage
Electric Vehicles could in principle provide large amounts of storage, and provide this, vehicle to grid (V2G). 30 million vehicles each with 50KWh of available battery storage could provide 1,500GWh of storage. There are however some issues to overcome.
The first issue is that car owners will not want to provide electricity to the grid if this reduces the life of the battery (it’s an expensive battery, as it’s optimised for weight and power). If the average 50KWh battery needs 50KWh per week (for a 300km range), then 1,000 charges should be enough for 20 years, which is about the lifetime of a car body. Only if the battery life is substantially greater than 1,000 full charge-discharge cycles (over 20 years), will owners accept V2G.
There are then some engineering issues around the grid not being designed for V2G, the need for DC to AC inverters, and the need for the vehicles to be plugged in whenever they’re not being used (as opposed to just at night, in the nuclear charging scenario). These can be solved, but the biggest issue is one of human behaviour. If there is a shortage of energy, vehicle owners will refuse to give their stored energy to the grid, unless they are 100% sure that they can get it back, and they can only be sure if there is standby capacity. Without the reassurance of standby capacity, they will horde energy, just like people horde fuel when there’s a refinery strike, and like people used to horde food in times of famine.
That means V2G can make no impact on the amount of standby capacity required, though it could play a role in reducing the usage of the standby capacity. We could probably assume that some – though not all – owners will join V2G schemes in return for lower tariffs.
Assumed Storage levels
How much storage would a 2050 renewables based supply model require? Any assumptions can be made, and Figure 7 shows the trade-off between storage and gas usage.
Further modelling assumes the following level of storage:
- 500 GWh from demand shifting – mainly of heat. This is only available when the daily average temperature is below 12C.
- 500 GWh of conventional storage. This could consist of:
• 200 GWh of V2G (10 million vehicles offering 20KWh each)
• 200 GWh of batteries – for example 10 million 20KWh power walls.
• 100 GWh of pumped storage
With 280GW of wind and 100GW of solar, this reduces gas usage to 12.8% of demand, as shown in Table 7.
Table 7 – Base case scenario with storage.
Adding tidal to the mix
Tidal power could in theory complement wind and solar as its output is not correlated. It is also reliable and predictable far in advance, which wind and solar are not. However, it is still intermittent in two specific ways:
- On a daily basis – the peak output from any installation will occur twice per day, with a period of zero output also twice per day.
- On a monthly basis – Spring tides are separated by a period of 14.77 days. Typically the neap tide is 40% of the spring tide, and therefore the extractable energy is close to the square of this – or 16%. In practice, it will be less than this as parasitic losses may be constant.
We assume that issue 1 can be resolved by adding storage. This is made feasible by two factors:
- The timing of high tides varies around the country.
- Tidal lagoons will provide peak power at low and high tide. Tidal flow will provide peak power at mid-tide.
The amount of tidal resource available is not yet known, and depends to a large extent on the costs that can be tolerated. If we assume a capacity of 40GW and a 18% capacity factor (as per the Swansea Tidal Lagoon Scheme), and, to slightly simplify things, that all spring tides are equal, then the daily output varies between 35GWh and 320GWh, with an average of 173GWh per day. This figure is in line with the tidal resource estimates by MacKay and others.
The extra storage required is impossible to calculate without knowledge of the sites and their distribution, and the mix of flow and lagoon, but is likely to be on the order of 1 hour of supply, or 40GWh. This is additional storage to the model.
Figure 8 – Daily electrical energy output from 40GW of tidal capacity. The average output of this is 7.2GW, or 2.7KWh per person per day. This is below the 5KW/person/day estimate of Mackay, but we need to consider what is economic. It produces 10% of demand.
Compared to wind power, tidal has a lower capacity factor – estimated at 18% based on the Swansea Bay data. It is also probably – though it’s early days – more expensive. However, it does have two advantages:
- Even at neap tide time, the 40GW of tidal produces 35GWh in a day. On a “bad wind day”, 40GW of wind capacity might produce 15GWh or even less (40GW x 1.5% x 24 hours = 14.4GWh).
- It is not correlated with wind or solar power (as far as we know). It is unlikely that “bad wind days” will coincide with a neap tide. (However, “unlikely” also means “possible” – this factor helps reduce gas burnt – but does not reduce gas capacity).
If we replace 40GW of wind power capacity with 40GW of tidal power capacity (and the extra tidal storage), and apply the outputs to the model covering four years of weather capacity, we see some marginal improvements.
Table 8 – Model output with 40GW of tidal capacity added to replace 40GW of (onshore) wind capacity.
Compared to the wind / solar solution alone, we see marginal improvements:
- Reduction in gas usage from 12.8% to 12.2% of demand
- Reduction in average carbon intensity from 80g/KWh to 76g/KWh.
- A reduction in spill from 34% to 28%
Figure 9 – Daily production, demand and storage figures based on the winter of 2014/15. Note that 1,000GWh of storage can be consumed in a day.
Given the marginal improvements, it probably only worthwhile investing heavily in tidal energy if it can provide electricity more cheaply than wind. On that basis, significant tidal is probably not viable even in a renewables world.
The preferred model
The renewables preferred scenario therefore consists of:
- 280GW of wind power capacity (or 200GW if offshore). This would occupy at least 20,000km2, and perhaps closer to 70,000km2 based on wind farm densities in the North Sea. The majority of this would be offshore, and the majority of that in the North Sea.
- 100GW of solar capacity. Hopefully, most of this could be accommodated on domestic and commercial roof space. If in fields, these would need to occupy about 2,000km2. (As with wind farms, most of the land can still be used for agricultural purposes).
- 500 GWh of electricity storage, or the equivalent of about 50 Dinorwigs.
- 500 GWH of thermal “electricity equivalent” storage, used as Demand Side response (demand shifting).
- A thermal fuel (gas, coal, diesel or biofuels) infrastructure with a capacity of 110GW, providing 12.8% of the electricity supply at a capacity factor of 9%. A large proportion of this is domestic / local scale fuel cells, providing combined heat and power.
The output from the model is reproduced here with a side by side comparison to the nuclear base case from Part 2.
Table 9 – Preferred renewables and nuclear scenarios side by side. The slight difference in demand is due to the fact the nuclear scenario was based on 20 years of weather data and the renewables scenario on four years.
We can also vary wind and solar capacities based on this storage amount, to see how this affects the shortfall that needs to be made up with gas.
Table 10 – Gas/Wind/Solar trade-off table assuming 500GWh of storage and 500GWh of heating storage
The emissions of 50.7 Mt compare with emissions from the energy supply sector of 153 Mt in 2014, which excludes emissions from nuclear, wind and solar. The carbon intensity of these sources is taken from the IPCC mid-range estimates and is highly contentious.
Indicatively, this would take up the following amounts of space:
Figure 10 – Indicative map showing the area occupied by the wind farms (in green) and solar panels (in yellow – though the panels would be spread over domestic and commercial properties)
Using hydrogen or synthetic fuels?
From Table 9 above, we can see that we are over-producing by 34%, which is currently wasted (when it’s windy in the UK, neighbouring countries won’t want more wind generated electricity). Can this be used to provide a “hydrogen economy” – either one based on hydrogen, or on synthetic fuels, thereby:
- Eliminating the need for fossil fuels for electricity
- Contributing surpluses to provide fuels for vehicles, aircraft and shipping
To do this, two requirements need to be fulfilled:
- (Over-production) X (round trip efficiency) > thermal demand. We see from Table 9 that our over production is about 217,000 GWh per year. To produce the 81,000 GWh of electricity that is from gas would need a round trip efficiency of 37.5%. That is not far off what is feasible – we may need to increase capacity by a small amount.
- (Electricity-equivalent storage) > (annual storage needs).
Figure 6 shows that if we have 35,000 GWh of “electricity-storage equivalent”, we don’t need to burn any fossil fuels. We can actually get away with less than that if we’re prepared to import fossil fuels in exceptional circumstances. If we’re not prepared to do so, then we actually need a lot more storage, as the consequences of running out of energy are severe.
A solution with hydrogen
The principle advantage of hydrogen is that it can provide large amounts of energy storage – in its liquid form. A 50,000 GWh store would require 2.1 million tons of liquid hydrogen, with a volume of 27 million cubic metres. Storing this volume of liquid hydrogen would be problematic to say the least – an optimal solution might be a network of over 400 tanks, each 50m in internal diameter, and storing up to 5,000 tons of liquid hydrogen. The heat gain would amount to some tens of MW across all sites and would probably not require any cooling beyond that required for liquification. Each site would probably need a source of water for cooling and warming the hydrogen. 5,000 tons of liquid hydrogen has the explosive potential of a small nuclear bomb, so the sites would need to be secure and remote. A possible location for many of the hydrogen storage facilities would be offshore, near to the giant wind farms that provide them with energy. This would also help reduce the transmission costs.
The principle disadvantage of hydrogen as an energy store is that it is inefficient. The process of making it by electrolysis, liquefying it, and then using it in a fuel cell to make electricity is about 36% efficient.
If we increase the amount of solar power capacity to 140GW, then our standard model now looks like this:
Table 11 – Production for hydrogen storage. The excess amount (“spill”) is converted into hydrogen, and then used to provide the “gas”. We have assumed a nominal 12g/KWh “excess” for electricity from hydrogen.
Evaluating storage needs on a 30 minute time base and starting with 48,000 GWh, the storage profile over a four year time frame is calculated.
Figure 11 – Storage amounts for a hydrogen store. GWh are expressed in terms of electrical energy, not chemical energy, assuming 60% conversion efficiency.
From this chart, it would seem a storage size of about 50,000 GWh would be required.
- Facilities are needed to convert the hydrogen to electricity at a rate of 105GW, and to electrolyse water at a rate of 180GWe (putting 110GW into the hydrogen store). In 2014 electrolysis uninstalled capital costs were estimated at US$400/KW, with replacement every 10 years. This implies a cost of about £6 billion per year for the electrolysis equipment.
- Storage of the hydrogen is a major engineering and security challenge, with 400 sites – mostly offshore – each storing 5,000 tons of liquid hydrogen.
- A hydrogen infrastructure has never been attempted at even a fraction of this scale. The construction and management of the storage facilities would be massive undertaking.
Based on this, if it is feasible to produce hydrogen from surplus renewables (or nuclear), then it would make more sense to produce synthetic hydrocarbons, using CO2 captured from sea water. The US Navy has prototyped the production of synthetic fuel, using hydrogen produced from electrolysis, and CO2 extracted from sea water. The Fischer-Tropsch process can produce methane or liquid hydrocarbons which can be stored easily in sufficient quantities to even out seasonal demand. It does however lower the overall efficiency of the process.
It should be noted that if we are using solar power to produce transportable hydrocarbons, then it doesn’t make sense to be doing this in the UK. The solar panels would perform better in the desert – and this could be any desert in the world. However, for the context of this document – a UK based renewables scenario – we will stick with 280GW of wind and 140GW of solar (perhaps we can “replace” 80GW of this solar with 25GW of desert solar). If we assume the efficiency of electricity to hydrocarbon conversion is 50%, and from hydrocarbon to electricity 60%, then our storage profile changes.
Figure 12 – Storage amounts for a hydrocarbon store. GWh are expressed in terms of electrical energy, not chemical energy, assuming 60% conversion efficiency.
Due to the lower efficiency, we have dropped a bit below zero. But that is not a major issue as there will still be natural gas available to make up a shortfall.
It would make sense to produce a variety of hydrogen and hydrocarbons, including:
- Hydrogen, for short term (up to a week) storage and electricity production.
- Compressed methane, for longer storage (up to a few months) and electricity production, as well as supplemental heating.
- Liquid fuels for long term storage and exports to the transport sectors.
A synthetic hydrocarbon storage system will require the development of a cost effective synthetic fuel production infrastructure, extracting huge quantities of CO2 from sea water, as well as approximately 105GW of electrolysis capacity. The cost and feasibility of this cannot yet be assessed. However, if renewables prove cost effective enough to make deployment of 280GW of wind and 100GW of solar feasible, then it would make sense to follow through and produce synthetic hydrocarbons using the excess production.