Designing a Carbon-Free Grid: an Explorable Explanation

In an earlier post I tried to understand whether the UK was windswept enough for wind turbines alone to provide all of its primary energy. The answer turned out to be: yes, as long as you don't think about intermittency too much. In this post I will be thinking about intermittency too much.

Understanding how much power you can expect to get, on average, from a renewable energy source is pretty straightforward, you just multiply your capacity by your expected capacity factor. It's tricky to get a good idea of intermittency without doing something very difficult like modelling the weather. But if we can lower our expectations from "good idea" to "rough idea" the task becomes a lot easier. Instead of developing a complex model of the sun, the wind and how each interacts with renewable infrastructure, then simulating what we think might happen lots of times, we can just look at historical data for what did happen with solar and wind grid contributions.

To keep things computationally simple I'm just going to look at one year of grid activity, and I've somewhat arbitrarily chosen 2019 as the subject year. So here's what grid supply and demand would have looked like in 2019 if we got rid of everything that wasn't domestic solar and wind:

Unsurprisingly this provides enough power to meet the UK's demand for exactly zero seconds of the year. Our task is to turn that on its head so that the UK has zero seconds without power.

But first let's look at what we've already got. I've stacked solar production (yellow) on top of wind production (blue) to make it clear just how close our sources get to meeting the demand line (red). As you'd expect, you can see quite clearly that solar is more productive in the summer than the winter. You can see less clearly that there is a bit more wind in the winter. This seems like a good thing. Each source should be able to make up for the other's seasonal deficiencies somewhat. Though the fact that demand is higher in the winter than the summer—thanks to the tendency of Britons to heat their buildings in the winter and swelter in the summer—means we will probably still want to skew towards wind.

Changing scale, if you zoom in so that individual days become clearer you can see that solar power peaks in the middle of the day and disappears completely at night. No surprises there. Demand is also strongly linked to the day/night cycle with a spike every morning as people get up for work and make tea, a slight trough while they're at work, another spike in the evening as they come home and make the other kind of tea, followed by a deep trough at night while most people are in bed. Meanwhile, at this scale, the wind more or less does what it wants.

Solar and wind weren't the only carbon-neutral sources of electricity the UK has. in 2019 hydroelectric power provided 1.8% of our electricity, while nuclear power provided 17%. "Other renewables" also provided 11.6%, but given that that will mostly be burning biomass I'm not sure how much I trust it to be carbon-neutral, so I'm going to exclude it from this analysis.

I'm excluding hydroelectric power too. <2% is a very ignorable amount of energy, and it's not an amount that we necessarily want to see grow for a couple of reasons. Firstly you have to be quite careful about where you put dams, if there's a lot of carbon in the soil that's being flooded then, once you flood it, that carbon will eventually be released as methane (a more potent, though more short-lived greenhouse gas than CO₂). This means a hydroelectric plant can actually be a worse greenhouse gas emitter than a coal power plant for the first 50-100 years of its life! The other reason is that there are really only so many places in the country that are suitable for dams, and pumped hydro plants; and most of them are either very beautiful, or very full of people's homes. Mount Snowdon, for example, might in theory make an excellent pumped hydro facility, but turning it into one would mean demolishing the visitor's centre and railway station we've already put at its summit!

So that leaves nuclear power: in 2019 the UK had 8 nuclear power plants with a total capacity of 9.15GW, though a rather underwhelming capacity factor of 63% meant that they only put out about 5.8GW of power on average. Let's take that 5.8GW number and shove it into our graph as a flat line; because you can sort of do that with nuclear power.

This is definitely an improvement. On windy, sunny days in summer, when demand is lower, there are points where supply almost meets demand! And no, I'm not counting the anomalous, sudden spikes downwards in demand. I'm not sure what they are, possibly power cuts.

But to turn "nearly, sometimes" into "enough, always" we're going to have to start cranking things up, which is very much what this post has been leading up to. If we work on the (dubious, but we'll go with it) assumption that doubling, say, wind capacity would take every point on the wind power plot and double its value then we can start to sketch out ways of decarbonising the grid. I don't think this is a terrible assumption. The UK has pretty disperse wind installations already, so assuming constant capacity, wind power at a given time is a rough measure for how windy the country as a whole was. So it stands to reason that if there were twice as many wind turbines around to catch that wind you'd get twice as much power; and I think this argument seems no less reasonable for solar power. For our over-simplified model of nuclear power it's definitely no less offensive than the original over-simplification itself.

So here's the same data visualisation as before, but with sliders that will let you add or remove capacity for each power source. For good measure I've included rough estimates for the capital cost of each power source, and for political reasons

I got the data for this project from Gridwatch, and realised that the mean values for both wind and demand were too low. For wind this meant that the capacity factor came in about 28% under the official government number for 2019. I emailed the guy who runs Gridwatch and he promptly emailed me back to tell me that the demand number doesn't include demand satisfied by embedded wind and solar, and 20%(!) of wind and all(!!) solar on the grid is embedded. He went on to explain that the wind number is wrong because government and industry have conspired to lie about the effectiveness of wind power, such that its capacity factor is actually closer to 20% than 30%. I sent a follow up email asking a few other questions like "Couldn't the unaccounted for wind be the source of this discrepancy?" but I never received a reply.

I'd naively stumbled upon someone with a serious axe to grind when it comes to renewables, one he's apparently been grinding for a decade . But, just because you're paranoid doesn't mean they're not out to get you. The pharmaceutical industry has churned out some truly miraculous drugs over the years, but they have a well documented history of publication bias, p-value hacking, and the like to make their crap drugs look worthwhile. By analogy I don't think we should assume that the renewables industry is above bending the truth for profit just because it's given us some handy windmills and solar panels.

So long story short: I've punted on this. If you're the sort of person who thinks the government puts fluoride in tap water for reasons other than dental health you can adjust the capacity factors down accordingly. If you're a gullible rube, leave them where they are.

I've included the ability to choose your own capacity factor within some reasonable bounds. It should be noted that capital cost is a pretty terrible way of comparing the costs of capacity. It doesn't include running costs or the expected lifetime of the installation. What you want for that is something like LCOE, but that's a bit trickier to calculate and compare for our purposes, so for simplicity's sake I'm including only capital costs and this disclaimer to give a very rough idea of relative costs.

One of the most remarkable things about this model is that if you take all three power sources and slam them as far right as they'll go (admittedly while leaving the capacity factors as they are) we still have a day without power, despite the fact we're generating 322% of demand on average! Even when we're producing over three times more energy than we need for the year, the intermittency of wind and solar mean that it's still not always being produced when we want it. The obvious solution to this is to store as much as possible of the energy that is over-produced, and then release that energy onto the grid at times when renewables aren't producing enough.

So here I'm going to commit more sins in the name of simplification, and assume that lithium-ion grid battery storage is the only game in town. This definitely isn't the case, but it is a relatively good all-arounder. Lithium-ion is already a good investment for short term (intra-day) storage (in part because grid batteries have a response time measured in milliseconds),and its improvement curve means as time goes on it should become viable for longer storage durations. Unlike pumped hydro, which needs to be built on a hill, batteries can be installed almost anywhere, including in buildings, so we don't have to ask any tricky questions like "does the country have enough viable sites for this to work?". So below I've included the same adjustable model as before, but with the ability to add battery storage. When there is an over-supply of power we can use it to charge the batteries as long as they're not full, and when there isn't enough power to meet demand we get power from the batteries if there's any available. Battery charge is measured in GWh shown on the right hand axis of the graph, note that the batteries only discharge to 15% as fully discharging batteries lowers their useful lifespan.

This is the final version of the model (for this post at least), and I recommend taking some time to play around with it to see what solutions you can find. Below are some solutions I've found, some discussion of the pros and cons of each, and whether I think they would actually work in practice.

Solution 1: Go Nuclear

If we keep the solar power we have, bump wind up to 18.19GW, and push our nuclear capacity and capacity factor as far as it'll go then we can get away without building any expensive battery storage. We would have to build 46 nuclear power plants (all of the UK's existing nuclear power plants are due to go offline by 2035) but this is doable. You generally want to put nuclear power plants on the coast because they need to be near a large body of water, a lower estimate for the UK's coastline is about 12,500km long, so we could have one every 271km, or double that if you build them in pairs, as we often do.

In his 2009 energy Bible Sustainable Energy - Without the Hot Air David MacKay called his nuclear heavy plan "Plan E" for "economics" because it was the cheapest plan he set out. 13 years later, and despite a freefall in the cost of wind, solar, and batteries this still looks about right. The cost of this grid is apparently around 13% of 2019 GDP which, spread over the next 29 years, seems like a very good deal.

I think, despite nuclear power's generally pretty good safety record this amount of nuclear power, and the waste it produces, is still, at best, a risk to be tolerated. This level of reliance on one power source is also inherently risky. The only feasible way of bringing this much nuclear power online this quickly is surely to churn out one or two reactor designs repeatedly and if there's ever an incident with one of these designs that shows it to have an inherent design flaw then the country would face a lot of very difficult choices.

Perhaps more damning than the hypothetical safety and reliability concerns is the fact that real world trend lines are not kind to this vision of the future. Globally, while the costs of renewables and batteries have been plummeting the cost of nuclear power has actually increased slightly. While the last few decades have seen massive growth for UK solar and wind, growth in British nuclear power sputtered to a complete stop in the 90s. There are some signs of a reversal to this trend, but even then, only one nuclear power station, Hinkley Point C, is actually under construction right now. Meanwhile we've closed two nuclear power plants since its construction began in 2018, and a total of four have closed since it was first proposed in 2010.

Overall this plan, or something close to it, doesn't violate the laws of physics or economics. Nuclear energy has proven itself able to provide reliable power to a far greater extent than intermittent sources plus grid storage ever have, so in some ways it's the safest bet. But it feels like it would require a far greater amount of political will to get this to work than other options, and I feel a great deal less confident in the viability of relying on that.

Solution 2: Just Enough Renewables

If we get rid of nuclear power, then provision just enough wind (81GW) and solar (75GW) to get average generation to just over 110% of average demand, we can then keep adding batteries until time without power falls to zero. This ends up requiring 21 days worth of battery storage. In contrast to the nuclear plan, this plan is very very expensive. Our sparing deployment of renewables costs "only" $192bn (7% of GDP), but because we only just have enough to manage we've had to spend $$3.8tn (135% of GDP!) on batteries to shift that energy around.

Because lithium-ion batteries discharge at a rate of about 2-3% per month the amount needed here represents a lower bound on the actual number, but I don't think there's any technical reason this couldn't work if you were able to source enough batteries. Being able to source enough batteries does represent a real, and probably insurmountable challenge though. The world produced 767GWh of li-ion batteries in 2020, this is projected to increase to 5,500GWh by 2030, but for this plan we need at least 15TWh! In reality, we'd need to turn to other forms of storage like hydrogen, compressed air, and pumped hydro to get this plan to work.

Solution 3: Over-provisioning Renewables

Let's start this plan by pushing wind and solar as high as they'll go: I've chosen an upper limit of just over 150GW for each, for pretty arbitrary reasons. Then, by keeping nuclear capacity the same as it is today (9.15GW, or 8 power plants), but pushing nuclear capacity factor to 90% we can get by with only two days of grid storage, at a total cost of $771.65bn/29% of GDP. A slightly modified version of this plan would be to cut grid storage in half, to one day, and to increase the number of nuclear power plants from 8 to 10. This would cut the cost to just $611.7bn/23% of GDP.

I'll be honest: I like this plan; in either incarnation. I think something like it represents the most viable chance at reaching net zero by 2050. It comes at a premium over the nuclear power plan, though I think the diversification of power sources alone makes that worth paying. It uses a lot of cheap wind and solar power, but doesn't require an unfeasibly large amount of land. More importantly it doesn't require as drastic a reversal in the UK's nuclear policy, nor does it need more batteries than the world is likely to be able to supply.

When the sun is shining, the wind is blowing, and the grid batteries are full, over-provisioning renewables will mean a lot of energy is going spare. Initially this might look like a bad thing, and when generation has to be switched off to protect the grid, it is. But it may also be this plan's biggest strength. Having huge amounts of intermittently very cheap renewable energy should provide an incentive for people to switch from on-demand fossil fuel energy to grid energy that's mostly used when it's cheap. This means having more electric cars and offloading excess grid energy into their batteries, rather than having cars we fill up with petrol when they're empty. In the UK, because a kW of gas is so much cheaper than a kW of electricity, it's cheaper to heat buildings and water using gas despite the fact that electrical resistive heating is greener and a 100% efficient process. In Norway the opposite is true because of their abundance of hydroelectric power. If the cost of electricity in the UK dropped to virtually zero every few days that calculus could be turned on its head here too! Eventually, intermittent over-supply could act as an incentive for investment in longer term storage so that solar and wind installations never have to pass on harnessing energy that's available to them.

Energy is a fractal of complexity, and to finish this post in an unreasonable yet finite amount of time I've had to make some pretty offensive simplifications. I've completely ignored a promising looking project that's currently underway to get solar and wind power sent straight from Morocco to the UK. It claims it will provide 8% of the UK's electricity, and should be able to do so more reliably than renewables here could. Ignoring biomass, which already provides huge amounts of energy, was probably over-zealous by any reasonable standard. I've bundled wind power up into one thing with one capacity factor and one capital cost, when in reality whether wind is onshore, offshore, or floating offshore makes a huge difference to both factors. Perhaps most importantly, lithium-ion is definitely not the only game in town when it comes to grid storage. These are just a few of the ways in which this model is far from perfect, but I do still think it's a useful tool for getting a feeling for the solution space.