Guest Post by Andy Dawson who is an energy sector systems consultant and former nuclear engineer.
How to decarbonise UK Power generation is a topic of heated debate, with renewables enthusiasts often keen to argue that there are a range of obstacles to the use of nuclear generation to meet more than a small proportion of total demand. Reasons cited are availability of space/sites, grid integration and the challenges of meeting variable demand. So, is an all-nuclear UK grid (with the small sleight of hand of pumped storage hydro in support) potentially viable?
I’ll set out an argument that it is indeed so, and more so that it comfortably exceeds any current carbon intensity targets.
The basic concepts arose from discussion on the website of the “Guardian” newspaper about the relative strength of fit between pumped storage on one hand, and nuclear or renewables on the other. That led me to do some basic numbers on how much pumped storage hydro (hereafter PSH) you’d need to meet UK daily demand variations on the assumption of a steadily generating nuclear fleet underpinning it. The first pass surprised me on how relatively close we were in terms of total PSH capacity (and in how few nuclear units basic demand could be supplied).
The other realisation was that the constraints on development of PSH are almost all to do with reservoir capacity (and particularly upper reservoir); there are few limits on the generation capacity associated with it (costs aside). If I’ve got an upper reservoir capable of storing sufficient water to give 20GWh of electricity production and a lower reservoir of greater capacity the site is technically viable whether I generate that 20GWh over 8 hours at 2.5GW, or 50hours at 400MW. There’s a secondary realisation in this area – the economics of PSH are crucially dependent on cycling the plant frequently and through the maximum possible proportion of capacity. So, it’s NOT good economics to have PS sitting for days or weeks with water in reserve.
What demand do we have to satisfy?
I’ve reviewed in detail the “Gridwatch” downloads available for 2011-12 at the level of (approximately) five minute reporting intervals. I’ve broken those into Winter (Dec, Jan Feb), Spring (Mar, Apr, May), Summer (Jun, Jul Aug) and Autumn (Sep, Oct, Nov). Recognising that designing for EVERY circumstance would be overly extreme, I’ve truncated these at the 2nd (minimum) and 98th (maximum) percentiles – since certainly in the upper bound case, having fossil-fuelled back up plant addressing peaking loads in such tiny amounts makes no material difference to overall carbon intensity. There’s also a strong economic argument for addressing peaking with plant with a cost profile leaning to (proportionally) minimal fixed and maximised variable cost.
These give (in MW):
Operating pumped storage in conjunction with a stable primary generating source is basically a matter of managing any differentials between primary output and demand by sending power to storage, or drawing it down. The generating (or pumping) capacity of the pumped storage. (i.e instantaneous MW) must be capable of delivering sufficient to fill the maximum deficit between primary generation and demand and ideally, to absorb something close to the maximum surplus. Obviously, over each period of deficit, the total energy available, adjusted for “round trip” efficiency must equal the total to be delivered over that time.
So, let’s look at those maximum surpluses and deficits; we can work on the assumption that we’ll size primary generation to match average demand and then define the surpluses and deficits on the basis of steady primary output to match that average:
Reviewing these two tables brings a few comments to mind:
- The demand minima vary less from season to season that gut-feel might have led one to expect.
- The delta between the 98th percentile and the peak is impressively large; however, this represents actually very little total production (on a rough calculation, in the winter period, and assuming that distribution of values above the 98th percentile is flat – a very pessimistic assumption – then it’d represent about 1% of total electricity production in the Spring period. Frankly, for the cost and carbon impact of that we may as well install or retain 10,000MW of CCGT and diesel and run it only sporadically.
- That the reasonably close match between average and median in each period means we’ll not be too far out in assuming a broadly symmetrical distribution of demand across the day – but this may give us an issue of arising from the need to have higher input energy into storage than the output we can withdraw due to round-trip inefficiency.
- And that the winter – perhaps unsurprisingly – gives us our “worst case” ; however, we need to modify that assumption slightly as we have a somewhat larger differential between absolute peak and average demand in Autumn.
The primary production issue
At the annual level, if I want to see an average production of 36.5GW, how much capacity do I need? 85% is a comfortable average capacity factor for a Light Water Reactor operating in baseload mode. That includes refuelling outages (for modern designs about 1 month every 18 months, although with advanced fuels that can be stretched to one outage every 3 years or so – 5.5%); it includes an assumption of unplanned availability of 2%, and to stay on the safe side, doubling the total of the two.
36.5/0.85 gives us 43.0 GW of total primary capacity.
Can we accommodate that in existing nuclear sites? All the eight currently licensed sites for new build, with the exception of Heysham are around 350-450 hectares – given that EDF operates six units on 150 hectares at Gravelines, we can treat those as unlimited in space terms. We’ll go through the details later, but 1300MW is an average size for a modern LWR – so that’s around 33 units, or an average of 4.1 per site. So at first glance, it’s possible. There are a few further constraints – for example, Oldbury is constrained in terms of cooling water availability, and generally you’d want to play a few games to balance load across the grid.
But let’s think about what might work. I’ll limit myself to existing designs (i.e. at least with a first site in construction globally, and known to be through, in or entering the UK GDA process), and assume that technologies currently earmarked for given sites are maintained. I’ll also assume a life extension to Sizewell B. That gives the following technologies:
I’m also assuming that the rumoured interest of China’s CGNPC and it’s partner CNNCin the currently undeveloped sites beyond Bradwell is real, and that the travails of the EPR are such that no developments beyond those currently planned will happen (actually, I’ve got my doubts even Sizewell will end up being developed with EPR technology…). That gives us:
Hinkley Point – 2x1620MW EPR currently planned. – 3220MW
Sizewell – 1x 1200MW SNUPPs unit (existing); 2x1620MW EPR currently planned. 4420 MW.
Wylfa – 2x1300MW ABWR currently planned. Double that up to 4x1300ABWR to give 5200MW.
Oldbury – 2x1300ABWR currently planned. Site constraints limit to 3x1300ABWR, so assume a total of 3900MW
Moorside – 3 x 1120MW AP1000 units planned. There are no space limits, but it’s a remote site relative to major demand therefore relatively unattractive to be very large so assume only one extra unit to give 4480MW
Bradwell – large site close to heavy demand. Assume CGNPC/CNPC interest materialises, resulting in 3x 1150MW Hualong-1 units – 3450MW
Hartlepool – large site. Assume CGNPC/CNPC interest materialises, resulting 2x1150MW Hualong-1– 2300 MW
Heysham – space constrained site. Assume CGNPC/CNPC interest materialises, resulting in 2x Hualong-1 units – 2300MW
Which gives us 29,270 MW.
We can add more units at those existing sites, but for grid reliability reasons it’d be best not to over-concentrate. Let’s set an arbitrary limit of four units per site, and assume any new development takes place as twinned units. That gives us a pair of Hualong-1 at Hinkley (given the EDF-CGNPC relationship), an extra pair of Hualong-1s at Hartlepool, and fourth unit at Bradwell – giving a grand total of a little under 35GW.
Beyond this, it’d be better to add some more sites. Where might those be? To start with, let’s rule out Scottish sites, given Scottish Government opposition to new-build nuclear.
The decision not to redevelop Dungeness was to all accounts marginal, and the site enjoys local support. Let’s add that option back in, but assume limiting it to a twin unit site given constraints imposed by the bordering Site of Special Scientific Interest.
For Grid purposes, it’d be good to have more capacity servicing the far west (Wales and the West Country), and another station somewehere close to the Humber Estuary, replacing the “megawatt valley” coal units of Drax, Eggborough and Ferrybridge.
For that former, the obvious sites would be next to the existing (soon to close) coal fired station at Aberthaw, or somewhere like Milford – although the LNG processing site isn’t the perfect neighbour. In either case, a sub-sea connector to Devon and Cornwall would be a useful addition, and improve grid reliability for South Wales and the West Country considerably.
For the Humber station, the most obvious spot would be next to the mothballed CCGT units at Killingholme in North Lincs.
In terms of technology, in order to balance the fleet it’d be best if the additional sites were a mix of ABWR or AP1000; that’d be 2240-2600MW per site.. Let’s assume a pair of AP1000ss at Dungeness and the other two sites as one twin AP1000 and one twin ABWR.
That gives total capacity of 42,540 MW from 34 reactors – roughly the same number that France deployed between 1973 and 1987. Only 500MW short of our 43GW target.
OK, what about the summer-winter variation? Well, that capacity/availability number was very much an annual average. By manipulation of refuelling cycles, we can reasonably assume close to 95% availability in the winter period. That gives almost exactly 40GW available – leaving us about 1GW short in winter. Or, one more site needed. Given heavy Thames valley demand, Didcot would be ideal, but be unlikely to be licensed, so I’ll suggest a pair of AP1000s at one of the Thames Estuary coal or oil sites – Isle of Grain or Tilbury, perhaps? The final grand total is:
Italics indicate not currently a nuclear site.
The map below gives an idea of the geographical spread of the sites:
And a breakdown by technology of:
A good spread of technologies, minimising risk of exposure to flaws in any given design, I think.
The pumped storage requirement
Our basic assumption will be that storage capacity is the constraint. We’ll also, for the sake of simplicity, assume that daily demand is roughly symmetrical around the mean (which is pretty close to the truth, in fact) and describes a sine curve.
In that winter period, for 12 hours a day, demand is greater than the average 41.1 GW. If it’s a half-sine wave over a 12 hour period with a maximum amplitude of 12.6GW that gives approximately 96GWh total demand over that period. On this we’ll be conservative – to keep a useful reserve for general system support, black start and so on, let’s add 15% – taking us to 110GWh.
What’s currently available? Installed and Consented/Planned Pumped Storage capacity is:
We’re about 32 GWh short. However, three other sites, all previously identified and evaluated for development can take us there:
With those in play, we have 127.2GWh of storage. Assuming these are reconfigured to deliver full capacity in either 5-6 hours (Dinorwig, Ffestiniog, Bowydd and Croesor) or 12 hours (the rest), it gives the following generating capacity (MW):
Giving 14.3 GW against that winter demand gap of 11.9GW. There’s also about 1.4GW of conventional hydro available.
Note that giving the Scottish storage units a longer operating time reduces loading on the cross-border interconnectors; allowing for about 4GW of local winter demand, that means about 3GW flowing south; night time MAY be more of an issue, with local demand, minimal basic generation, and up to 7GW flowing north just to recharge the storage. We may need to beef up the interconnectors up to 9GW or so – or add Torness and Hunterston back to the list of new build sites…
So, where does that leave us in the grand scheme of things?
In terms of system capacity – 44.8GW nuclear, 14.3GW of pumped storage, 10 GW of (infrequently used) CCGT and diesel, and 1.4GW of conventional hydro, to give a grand total of 70.5 GW. Against peak demand of 59.2GW, that gives a system reserve of 19% – a far, far better place than we are in at the moment.
In terms of carbon intensity – roughly 98% of basic production is nuclear and 2% CCGT/diesel, which on the IPCC figures of 11g/KWh and approximately 450g/kWh respectively gives just 20g/kWh – comfortably beating the Climate Change Committee’s basic assumption of 100g/kWh and stretch target of 50g/kWh.
In terms of cost, estimates are harder to make; all of ABWR, AP1000 and Hualong show every indication of being considerably cheaper than EPR – so let’s assume a conservative £75/MWh (the aggregate strike price on offer if EDF proceeds with Sizewell C values output from that plant at £86.50/MWh). Some rough calculations on pumped storage undergoing heavy cycling (perhaps the subject of another post…) give something about £30/MWh incremental cost for every unit that passes through them. Gas and diesel costs will be considerably elevated from where those technologies currently sit, given impacts of the need to amortise fixed costs over a small production volume. A working assumption of £150/MWh would seem reasonable. That gives an aggregate cost in the order of £80/MWh.
Now, in one sense, this is only a partial analysis. To be done properly, it needs multiple years of data, and much closer examination of issues like the controllability/variability of demand within the daily cycle and the matching of production to it. There’s also a minor task in validating that the surplus power available at times of low demand matches adequately with the necessary inputs allowing for “round trip” inefficiencies. However, I don’t believe at this stage that those will shift the picture materially, and the “nuclear option” is at very least technically viable.
The bigger issue is that to meet grander decarbonisation objectives, we’d need to factor in both demand growths and considerable changes to the daily/seasonal demand variation. That needs to be the subject of a much larger, and rather more speculative analysis exercise.