top of page
  • Writer's pictureGeoff Russell

Is nuclear power really expensive? or is it just better?

[This is Part I in a series of posts on nuclear costs, they are companions to the post on nuclear construction times. Some of it is written from an Australian perspective, but most is more general. ]


  • AEMO: Australian Energy Market Operator

  • CCS: Carbon Capture and Storage

  • EV: electric vehicle

  • ISP: Integrated System Plan

  • NREL: United States National Renewable Energy Laboratory

  • OWID: Our World In Data

  • PV: photovoltaic

  • RE: Renewable Electricity

  • RIS: Renewable Integration Study: Stage

  • UNECE: United Nations Economic Commission for Europe

  • VRE: variable renewable electricity

US renewable experts on lowest cost grids

In a June 2021 scientific paper, lead author Paul Denholm and sixteen other researchers from three US renewable energy institutes said something that would surprise many Australians:

Many studies suggest a mix of low-carbon resources including nuclear and Carbon Capture and Storage (CCS) could achieve decarbonization at the lowest costs, and it is important to evaluate the merits of 100% Renewable Electricity (RE) [Renewable Electricity] systems compared with alternatives that achieve societal goals of addressing climate change.

Note the crucial points: many studies, nuclear and lowest costs.

Who were the institutes?

  1. The United States National Renewable Energy Laboratory (NREL),

  2. the Office of Energy Efficiency and Renewable Energy in the Department of Energy, and

  3. the Renewable and Sustainable Energy Institute at the University of Colorado.

So why do so many Australians, including our Prime Minister, Anthony Albanese, for example, say that nuclear is the most expensive form of energy? Because he’s talking about the “levelised cost of electricity”.

What’s this?

The levelised cost of electricity is like the levelised cost of calories. Every food has a caloric value; the energy you can get out of it when you eat it. If all you are worried about is calories, then you can compare tofu, bacon, sugar, bread and soy milk and calculate the cost per calorie. Sugar is the cheapest.

But the minimum cost balanced diet is a very different thing from the cheapest source of calories. A balanced diet needs fat, protein, carbohydrates, minerals, fibre and vitamins; and a large number of phytochemicals is beneficial also! So it is with electricity grids. There is a bunch of things grids need with technical descriptions like frequency stability, voltage stability, rotor angle stability, power system protection, voltage control and the ability to start when the entire grid has collapsed (known as “black start”). Some of these attributes are normally grouped together and called “system strength”. But system strength doesn’t have a good technical definition, it’s just the ability of the system to keep a steady voltage in the face of stresses, like things falling on power-lines.

Wind and solar provide none of these grid nutrients; they are just cheap calories.

The irony is that the value of these essential grid nutrients wasn’t realised when they were provided for free by thermal power plants (nuclear, gas, coal and hydro). As thermal plants shut, and these nutrients are in short supply, they are worth serious money, which is making nuclear look very valuable and easily worth the high price tag attached in countries with broken regulatory systems.

These nuclear advantages now show up in proper modelling as acknowledged by the US renewable energy experts in the quote at the beginning of this post.

Before proceeding, it’s worth describing power system protection; it’s about handling failures. When things fall on power-lines, which they do every day, you need to both find the fault and shut off the power. What happens in a strong grid is that the location of such faults is detected by the surge of current that occurs. The surge trips one or more switches which isolate the area and shut of the power. It sounds so natural, but it isn’t, it’s brilliant engineering and really hard to do when you have current flowing in both directions and renewable sources that disconnect when they detect changes in frequency or voltage. In essence, that disconnection, to save their own arses, causes a bigger problem but doesn’t provide the current surge that will help with the fault location and isolation. The result is that faults are harder to find and fix.

Power system protection used to be provided (for free) by power plants and the large hunks of spinning metal in their generators.

Renewables also don’t provide operational reserves; meaning generators that will do exactly what they are told to do; and when they are told to do it.

The opening quote above talks about the “merits” of 100% RE. So what are they? The paper doesn’t say; which is because the paper is all about the challenges of such systems, hence the title: The Challenges of Achieving a 100% Renewable Electricity System in the United States.

Academic language can be a little obtuse. Here’s my suggested paraphrase of that preceding quotation in plainer English: Building electricity systems with up to 80% renewables should be relatively easy, but beyond that, it starts getting really hard and very expensive; it will be cheaper and simpler to finish off with nuclear and/or gas with CCS. So people advocating 100% RE grids will need to come up with good reasons (i.e. merits).

The paper then lists challenges in getting to 100% renewable electricity in the US; and it’s a bloody long list. It’s too long a list for any one paper, so they decide to focus on the techno-economic issues and ignore:

“… challenges of siting RE and associated transmission, land use, environmental and aesthetic issues, material supply and manufacturing scale-up, grid security, and considerations of equity and social justice.

Fair enough. I’m also partial to ignoring the big hard nebulous problems to focus on the relatively well defined technical details.

Consider grid security; by which I mean cyber security. Renewable advocates are quick to praise virtual power plants and selling to the grid but, like the US experts, like to ignore the downsides of having every building and vehicle in the world addressable from a hacker’s phone anywhere on the planet.

On the other hand, maybe I’m being selfish. Think of the new jobs it will create. Hackers will be able to upskill from sending spam and hacking fridges and baby monitoring devices, to hacking virtual power plants, electric vehicle (EV) charging stations, and via them, EVs. As we speak, the Kremlin is targeting wind farms.

One of the key challenges to 100% RE systems is, as the US renewable experts point out, costs.

Why costs, if renewables are so cheap?

Our dietary analogy is too crude to capture the problem in its entirety.

The US experts point out that costs are non-linear as variable renewable electricity (VRE) (wind and solar) penetration gets past about 80%. The word “non-linear” used without details and equations, serves two purposes in this context. First, it tells you that getting from 85% to 90% RE is harder and more costly than getting from 80% to 85% RE and getting from 90% to 95% RE is harder and more complex again. Second, it acts like a lighthouse warning about hidden reefs and rocks which can rip holes in the best laid models and plans of modellers.

I did a post recently about grid modelling: “Modelling grids and telling fibs”. It split models into three kinds. Two of the three kinds are (though I didn’t mention it) linear. Linear is “easy”; meaning all you need is a maths degree and years of practice. You can have thousands of equations and even more variables and computers will spew forth answers. Once things are non-linear, all bets are off. Even small non-linear problems can be really hard to solve. Mostly, what people do is try and find a way of using linear equations to approximate the non-linear ones.

Think about a straight road, you can predict in your mind where it is going without having to travel along it. And two straight roads? You can predict with high school maths where they will intersect. You can do this with thousands of straight lines using a computer. Now think about a windy road. How to predict where it goes? Short of a road sign, you have to travel along every metre of it. How to predict where two windy roads intersect? You can’t; not without travelling along them. Some windy roads can be approximated by straight segments; and then you can apply the machinery of straight lines.

In the modelling post, the 3rd level of models were of physical networks. The reason they are hard is because they are non-linear; not arbitrarily windy, but not straight. Despite involving nothing more than well understood physics, handling such models remains challenging as network size and complexity rises; even with best modern computing resources.

Is there anything worse than trying to model large collections of non-linear physical devices?

The US researchers reckon there is; namely, large collections of inverters (the things that connect solar panels and wind turbines to the grid). This is because inverter behaviour isn’t just a matter of physics, but of the inscrutable choices of geeks writing computer code.

Suppose someone upgrades the software (remotely) on a bunch of inverters without telling the grid modellers. Suddenly, your best physical models may be giving you the wrong answers; but you won’t know until some prediction fails. At the physical level, this kind of problem is already biting as multiple sets of inverter standards roll out in a frantic effort to get the rules right. In May 2021, Australian Energy Market Operator (AEMO) reported on inverters simply not working as specified. They also reported that many inverters built to older standards (meaning 2005 and 2015) didn’t work as required by either those old or the latest standards: AS/NZS4777.2:2020. We are talking about 30 to 50 percent of devices not working as they are supposed to. The consequence? “Extensive disconnection of DPV [distributed photovoltaic (PV) … households] in response to voltage disturbances”. This can mean that small problems can snowball into larger ones. This is already impacting standby power and frequency reserve requirements. It makes modelling transient electrical behaviour, already diabolically difficult, even harder.

When you only had a few large generators, you could easily ensure that they functioned as required, but a large collection of mass produced consumer grade electronics? Not so much. You can read the US study for their collection of challenges, or read AEMO’s own specification of the problems, in its Renewable Integration Study: Stage 1 (RIS). Don’t forget to read the appendices, this is where most of the critically important stuff is hidden.

Our AEMO appears to agree

The US study related to the US electricity system, where about 20 percent of electricity is already provided by nuclear power, but their judgement on the costs and feasibility of 100% RE grids is consistent with our AEMO’s approach in its 2022 Integrated System Plan (ISP).

The ISP is not a 100% renewable electricity plan, but instead of nuclear power, 14% of electricity will come natural gas; which it says is for “firming and peaking”. These gas based emissions in the ISP must be offset. The operating emissions of natural gas depend on the efficiency of the generators but are something like 412 grams per kWh, implying a carbon intensity for the ISP grid of 58 g-C02/kwh … over and above the lifecycle emissions of building and decommissioning or recycling the equipment (solar farms, wind turbines, batteries, transmission lines and so on). AEMO’s ISP ignores the construction costs, but these should be minimised and offset. Ignoring this is as silly is only considering the tailpipe emissions from an EV and ignoring the costs of building the vehicle itself, especially the battery.

It seems clear that AEMO, like the US experts, didn’t think getting that last 16 percent from renewables was worth the hassle or cost, but they had no choice but to opt for fossil fuels because the Greens and a sizeable portion of the ALP have long preferred fossil fuels to nuclear power. If we got the last 16 percent from nuclear then the operating emissions would be zero and the life cycle emissions would also be lower.

In short, these US renewable energy experts acknowledge that the lowest cost path to zero emission electricity includes nuclear (or other dispatchable fuel based source like gas with CCS). How can this be true if nuclear is so expensive? Read on.

But first …

Who supports nuclear power in Australia, and why?

Australian opposition leader Peter Dutton’s calls for a consideration of nuclear power in Australia have prompted a flurry of fustian rhetoric on the topic; with new Prime Minister “Albo” enjoying poking vacuous fun at the idea. Anthony Albanese, who pledged to do politics “differently has flunked out at the first opportunity to walk the talk. Sadly, such is the political bestiary in our parliament, that any policy of your enemy is shite by definition.

No-one loves a reactor like a reactionary [applause from the peanut gallery] (Albanese ... Hansard, 3/8/22)

Fortunately, many scientists aren’t swayed by bullshit rhetoric and think clearly about energy sources and there is a particular group with both better rational thinking skills and much higher environmental credentials than either Dutton or Albanese, not to mention the Greens.

Back in 2014, the then Professor of Climate Science at Adelaide University, Barry Brook (now at the University of Tasmania), put together a list of scientists who signed an Open Letter to Environmentalists on Nuclear Energy. The list of 75 scientists is a veritable who’s who of ecologists and biologists working in Australia. It isn’t surprising that scientists, people who evaluate evidence for a living, should think differently from politicians.

Why not? Because a 2022 Life Cycle Analysis of electricity technologies by the United Nations Economic Commission for Europe (UNECE) confirmed earlier studies when it scored all the major electricity technologies on a large set of (mostly environmental) attributes. They found that, compared with wind and solar, nuclear power used less land, was less eco-toxic, used fewer mineral resources, and produced fewer greenhouse gas emissions over the entire life cycle (including waste handling and decommissioning) … see here for the long list of attributes and the relative scores of nuclear and renewables.

In short, nuclear power has a much lower eco-footprint than wind and solar. It also doesn’t need batteries and a grid redesign and rebuild.

Until we have nuclear fusion, nuclear fission is simply the best technology we have; with daylight rather than sunshine for second.

Arguments about nuclear power used to focus on radiation and safety. Increasingly, people have realised that these arguments are fallacious and often disingenuous. Our World In Data (OWID) is one of many good quality sources that have looked at the data and presented the results in a clear and simple form; nuclear power is among the safest of energy sources.

It’s a bit like the cars vs planes "issue". Planes are safer than cars despite the intense fear some people have of them. But the analogy isn’t perfect. Planes crash and kill hundreds of people from time to time; nuclear power has never done this. I’d welcome detailed suggestions from any reader on how this could happen; I can’t think of a way.

But what about the costs? And aren’t they slow to build?

This leaves us with two remaining issues: cost and construction speed. Waste, in case you were wondering, has always been a non-issue; demolished by even the most basic of knowledge about radiation plus a little thought. See here for details.

I dealt with construction speed in a companion post here. Put simply, the fastest national per person electricity generation roll-outs in history have all been either nuclear or based on fortuitous geography (hydro and geothermal).

There are two aspects to nuclear costs, the stand-alone cost of building and operating a reactor and the impact on grid costs of having nuclear reactors in the mix.

END PART I Next post … Nuclear construction costs



bottom of page