Don't switch off clean power

Britain's 2030 Clean Power goal depends on keeping 2 1980s nuclear plants running

Britain used to be a nuclear superpower. In 1932, the atom was first split in Britain. In 1956, Britain opened the world’s first full-scale commercial nuclear reactor. Less than ten years later, it had built 21 more. As late as 1965, Britain had more nuclear reactors than the rest of the world combined.

Yet Britain hasn’t completed a new nuclear power station in almost thirty years and most of our remaining fleet is set to be taken offline in the next few years. Only Sizewell B, which opened in 1995, is planned to stay online past 2028. Assuming one unit at Hinkley Point C has not been completed by 2029 and Sizewell B will close for two months then Britain will have no nuclear power whatsoever on the grid. For the first time in more than seventy years, the sun will set on British nuclear power.

For energy security, household bills, and net zero, this situation must be avoided. We can and should extend the life of our existing fleet of Advanced Gas-cooled Reactors (AGRs) to avoid it happening.

What will take nuclear’s place on the grid? Some of the time, it will be renewables like wind and solar. Batteries charged at times of high wind or sun will pick up some of the slack too. But there are limits to intermittent renewables and the short-duration batteries they charge. When the sun isn’t shining and the wind isn’t blowing, our grid will more often than not fall back on expensive, carbon-emitting natural gas. The result will be higher emissions and higher electricity prices.

The planned phase-out of Britain’s remaining AGRs therefore threatens to derail the Government’s Clean Power by 2030 target. This is the finding of the now-Government-owned National Energy Systems Operator, or NESO’s new report into the feasibility of the Government’s “Clean Power by 2030” target.

Both of NESO’s pathways to Clean Power by 2030 rely on us having between 3.6GW and 4.1GW of nuclear power online. In these scenarios, while nuclear’s contribution to the grid falls, we still manage to keep one AGR online (either Heysham 2 or Torness) as well as bringing one of Hinkley Point C’s units onto the grid.

If we fail to keep even one of our remaining AGRs online, then we will be left with just 2.7GW of nuclear capacity, pushing up the amount of unabated gas we burn by as much as 11TWh. This would breach the NESO’s ‘less than 5% unabated gas in a normal year’ definition of Clean Power. In short, it will mean mission failure.

The more nuclear we can keep online or add to the grid, the less gas we have to burn. Britain has four AGR power plants left, all built during the 1970s and 80s. If Britain can keep even two online, along with getting one of Hinkley Point C’s reactors online and keeping Sizewell B running, then Britain’s nuclear capacity will be 5.3GW, or 34TWh of clean power each year.

NESO’s modelling shows that keeping these two AGRs alive – or alternatively, replicating lost capacity through fast-built Small Modular Reactors – would allow Britain to keep the amount of unabated gas burnt to just 3.3% (11TWh) of our total power mix. This would reduce carbon emissions by 8.5 million tonnes – the equivalent to taking 1.8 million petrol cars off the road. Depending on gas prices in 2030, retaining at least two AGRs would cut annual household bills by between £13 and £95 (most likely estimate: £35).

Caption: Note - Heysham 2 (Capacity: 1.25GW) and Torness (Capacity: 1.29GW) have similar capacities to one unit at Hinkley Point C, but due to age and maintenance requirements have a lower load factor. Cost savings use CP2030’s gas price assumptions. In addition, Low savings reflect diminishing returns from additional nuclear beyond NESO’s 3.5GW baseline. Central and High savings reflect likely additional demand/supply shortfalls.

You may have noticed that the table above uses two separate numbers for the amount of gas displaced in some scenarios. The lower figures assume electricity demand is constant and every other requirement necessary for Clean Power by 2030 (grid, renewables, demand flexibility, and CCUS) is met. In such a scenario, diminishing returns from additional nuclear generation set in quickly. In this scenario, if nuclear capacity exceeds 4.1GW, each additional TWh of nuclear only displaces 0.25TWh of gas. Most of that extra nuclear output just leads to renewables being constrained because supply exceeds estimated demand. But is that scenario likely? As NESO put it, reaching Clean Power by 2030 is achievable, but requires a herculean effort on multiple fronts. There are multiple routes to failure:

• If we fail to build enough pylons connecting new renewable supply to where the demand is, then there will be a shortfall that will have to be filled by gas. 

• If we fail to get all of the renewable projects and carbon capture projects we need approved by the planning system in time, then there will be a deficit of clean energy for gas to make up. 

• If there are more delays at Hinkley Point C and we can’t even get one of its units online, then – you guessed it – we will need to burn more gas to keep the lights on.

And that’s just the supply side. There is a high risk that we miss the Clean Power by 2030 goal solely down to unexpected growth in power demand. NESO estimates that by 2030 data centres will demand 22TWhs worth of power annually. This is a big leap from today – where just 5TWh of power is required for data centres – but it is still likely to be an underestimate of 2030’s power requirements.

NESO base their assumption of 2030’s data centre demand from registers of grid connection and planning applications. But it is surely risky to assume that no further data centre projects come forward during the next few years. 

Is there an implicit belief in NESO’s modelling that any new power-hungry data centre project would get rejected at planning because it’s not compatible with the Government’s Clean Power goal? There’s a hitch in that plan. Our Government seems to want more, not less, investment in data centres. It commissioned entrepreneur Matt Clifford to look to come up with an AI opportunities plan. The FT reports that Clifford’s plan will propose special planning zones where new data centres and complementary energy infrastructure can be built without going through the usual bureaucratic planning process. If the Government adopts Clifford’s plan then NESO’s forecasts for data centre power use are likely to require urgent revision.

It wouldn’t be the first time that NESO has had to raise their data centre demand forecast. As recently as 2022, they were forecasting 15TWh of demand in 2030. The new forecast is almost 50% higher, at 22TWh. Another similarly sized revision to 30TWh would create an 8TWh supply shortfall per year. Renewables that we are currently paying to not produce may be able to cover some of that demand shortfall, but a large chunk, say four-fifths, will come from burning gas.

One way for the Government to think about AGR extension is as an insurance policy for the Clean Power by 2030 target. Every additional GW of nuclear capacity on the grid gives us more options if we fall short of the main plans due to slow grid buildout, planning delays for renewables, or higher than expected data centre demand.

What will it take to extend our AGRs?

Safely extending the life of nuclear power plants far beyond the date they were expected to close is normal practice across the world. In fact, we're doing it right now in the UK. Sizewell B, the last nuclear reactor Britain built, was expected to last 40 years. That was 29 years ago, but EDF is planning for it to stay operational all the way up to 2055.

88 of America’s 92 nuclear reactors have received 20 year extensions. 15 reactors have applied for further extensions, which would give them a potential lifespan of at least 80 years.

EDF’s reactors in France have already received licences to continue operating for up to 50 years and their senior executives believe extending them to 80 years is achievable.

In the US, Microsoft is working to re-open Three Mile Island in order to power a data centre. In Germany, leading opposition politicians who have come to see their nuclear phase-out as a costly (and environmentally destructive) mistake are making the case for reopening nuclear power stations.

One study from 2023 estimated that restarting each German reactor would cost between €100-200m and take about a year. If Spain and Belgium followed suit, then Europe would have enough power for 100 of the world’s largest AI data centres at ultra-low (€25MWh) prices.

Could Britain’s nuclear fleet be extended at similarly low cost? Britain’s Advanced Gas-Cooled Reactors have been extended before. Although they were originally planned to operate for 30 years, many AGRs have lasted longer. Torness, Scotland’s most productive clean power asset of all time, will have lasted for 40 years if it closes as planned in 2028. Hunterston B, which held that crown until this month, lasted 48 years. Extension is possible, but AGRs present special challenges.

One problem is the design of the UK’s Advanced Gas-Cooled Reactors. Britain went in a different direction to the rest of the world when it came to reactor design. In theory, AGRs had the potential to be massively more efficient than other designs because they could be refuelled without being shut down. However, their construction was a case study in British infrastructure failings.

In a fascinating post on why nuclear flopped in Britain, Progress Ireland’s Fergus McCullough lists some of the problems:  “no prototype had been built, the very designs themselves made construction challenging, the materials hadn’t been tested, and the stations were too dissimilar to allow modular construction (which would have saved money).” 

Trying to go alone and develop our technology – instead of adopting cheaper US designs that actually had a track record of working – created and is still creating unique challenges. 

AGRs have a number of key flaws that limit how long they can safely (and efficiently) operate. It is sadly extremely unlikely they can make the 80 year lifespans that some US-designed reactors are set to achieve. 

One key problem is the inability to replace the graphite core. Over time, cracks form in the graphite core. If enough cracks form, it may prevent the reactor from safely shutting down in the event of a major natural disaster. Due to the way AGRs were designed, it simply isn’t possible to fill in the cracks. So to allow operation to continue, the Office for Nuclear Regulation must be persuaded by EDF that the cracks do not risk disaster. The key regulatory concern is whether or not the reactor could still safely shut down in the event of a 1 in 10,000 year earthquake.

It should be noted, by the way, that even in the unlikely scenario the AGR is hit by an unprecedented earthquake, AND more than four-fifths of the control rods can’t enter the reactor to cool it, AND the backup cooling Boron injection fails, AND the other backup Nitrogen injection also fails, it is still far from clear that a single person would be hurt (let alone killed). 

The reactor may melt down, but it would still be contained inside a massive concrete containment cell. This is, let’s be clear, far from ideal and will probably be a very expensive problem, but the public (at least, the people not killed in this hypothetical massive earthquake) can sleep safely at night.

The issue is less the question of whether the reactor could shut down in the event of a 7.2 earthquake (10 times more powerful than Britain’s worst earthquake) and more the question of how to prove it. Proving it is really, really, expensive. Not only does it require updating a complex modelling tool called GCore, but it also involves investing in physical lab tests (like this project at Bristol).

It should be noted that these aren’t the only issues with AGRs. There’s also the small problem of the fact that the boilers are fully encased in concrete and thus inaccessible. EDF will likely be reluctant to invest heavily in safety modelling, if there’s a risk that they have to shut down operations for a completely different reason altogether. This boiler issue may rule out the older two reactors from life extension to 2030 – it is a matter of when, not if – but experts I’ve spoken to suggest the younger Heysham 2 and Torness could keep going up to 2032/2033 before this becomes an issue.

Let’s not repeat our past mistakes

We’ve been here before. When Russia launched its illegal and unprovoked invasion of Ukraine, gas prices spiked. Britain faced a genuine risk of blackouts as the cost of securing gas supplies surged. Coal plants set for closure were kept open. Parliament even debated and voted in favour of bringing fracking to Britain. The Government published the British Energy Security Strategy, which included massively expanded targets for nuclear, wind and solar. And yet Britain allowed one of its AGRs, Hinkley Point B, to close. 

Over its 46 year lifetime, Hinkley Point B was Britain’s most productive clean energy asset providing enough power to boil the equivalent of 100 billion kettles. In 2021, it met 3% of Britain’s total power needs. It produced no greenhouse gases in operation and hasn’t harmed a soul. Yet we let it close at the worst possible time.

In a winter when energy prices reached all-time highs and the Government committed billions of pounds to keep them at merely eye-watering levels, we let 1GW of reliable clean power leave our grid. It was an expensive error. As a result, Britain was forced to replace Hinkley Point B’s 10TWh per year with alternative sources – imported gas and coal. If we assume Hinkley Point B could have been kept at a cost of £45 per MWh – if anything, a conservative estimate – then in the Winter of 2022 alone Hinkley Point B’s closure cost billpayers and taxpayers around £600m.

Hinkley Point B’s closure wasn’t inevitable. It was a policy choice. One of the worst policy choices then-Energy Secretary Kwasi Kwarteng made in 2022. His department washed their hands of the choice. They described it as simply a matter to be dealt with by EDF and the Office for Nuclear Regulation. No written request from the Department was made to EDF. 

At the time, one energy policy expert told the Guardian: “On Hinkley B, there is no technical reason it can’t continue to operate, if they can satisfy the regulators. For EDF, it will be fundamentally an economic decision.”

What EDF needed more than anything was confidence that if they went through the expensive effort of producing a detailed safety case for extension explaining what would happen in the event of an unprecedented huge earthquake in Britain, they wouldn’t get shut down regardless for some other reason. 

Under the status quo, politicians have outsourced decision making to regulators. To be clear, I don’t think it would be appropriate or useful for Ed Miliband to inspect nuclear reactors. But, as it stands, the Office for Nuclear Regulation not only inspects whether there is a risk of a fatal nuclear accident, they are also responsible for deciding what we do about that risk.

Without democratic oversight, there’s a danger that we fall back on uninterrogated regulatory principles with costly consequences. For example, one reason Sizewell B is set to be down for maintenance so frequently are the countless safety features insisted on by the regulator. Their view was, and still is, that a safety change was only grossly disproportionate when the costs outweigh the benefits by a factor of 10. In other words, their view is that a safety measure that prevented £1 worth of damage is worth spending £10 on.

When Britain was in the grip of the Covid pandemic, Sir Patrick Vallance and Prof Chris Whitty repeatedly emphasised that their job was to advise and present the science. It was the responsibility of our elected government to decide whether or not the costs of closing schools and businesses were worth the benefits in cutting the disease’s transmission.

We need a similar approach to AGR extension. Our expert regulators should present their best assessment of the risks of AGR extensions. And our elected government should decide if the benefits of delivering Clean Power by 2030, displacing 6 billion cubic metres of gas, and saving households up to £134 each year mean they are risks worth taking.