The end of "studies show"

I've built a deterministic simulator of the British electricity system. It replays the last forty years of real weather, half-hour by half-hour, against any grid you care to describe.

AS MY BOOK “The Energy Trap” heads towards publication, I’m starting work on its followup. This one will be called “The Grid: A User’s Manual”. It will be an explanation, in the jargon-free style of “The Energy Trap”, of how our electricity grid works — and increasingly malfunctions.

The books are natural companions. “The Energy Trap” describes in general terms why weather-driven electrical grids can’t work physically, technically, or economically. “The Grid” describes why for this particular one.

Britain is currently rebuilding its electricity system around the weather. Over the next two decades, most of the electricity used in this country is intended to come from occasional wind and sunshine, and the demands placed on that electricity dramatically increased: heating that used to burn gas, and journeys that used to burn petrol, are to be powered from the grid instead.

Whether this plan works — and what it costs — turns on a small number of questions that sound simple and are not:

• How much backup does a weather-driven system need, and of what kind: power stations held in reserve, stored energy, or cables to neighbouring countries?

• How much energy storage is enough? Enough for a calm evening, a still fortnight, or a poor decade?

• What happens in the worst weather this country actually experiences — not an average winter, but the coldest, stillest weeks in living memory, when heating demand is highest at exactly the moment wind supply is lowest?

• Does the grid remain stable, second by second, when the heavy spinning machinery that has steadied it for a century is retired?

• And what does each version of the future system cost, per unit of electricity actually delivered to the people paying for it?

These questions, why they matter, and the problems that come with them, are the spine of “The Grid”. Understand these, and you’ll understand better than most how our grid works .

But, as I observe in “The Energy Trap”, what passes for energy policy now is being devised by those with exciting qualifications such as “Public Policy”, “Political Science”, and “Greek Classics”, who suppose that these questions are matters of mere preference. That they can be decided, in other words, by how we feel about matters such as “the climate”, “social justice”, and “lifestyle preferences” (i.e. theirs, not yours).

These question are in fact quantitative, and constrained by brutal physical and engineering limits that can’t be settled by principle, preference, or slogan — only by arithmetic done carefully on the right data.

Such arithmetic as is done rests on methods with known blind spots. Studies that test a proposed system against a single year of weather cannot see the rare events that set the true requirement, because the worst week in forty years is, by definition, unlikely to be in the year you picked. Annual averages hide the hours that matter: a system can be in surplus for 8,700 hours of the year and still fail catastrophically in the other sixty. And the cost measure most often quoted in public — the lifetime cost of one wind turbine divided by its own output — says nothing about what it costs to make that turbine’s output useful: the backup, the storage, the spare capacity, and the wires that turn intermittent generation into reliable supply.

The arithmetic fails, in other words, because it doesn’t account for how the electricity grid actually works, over the long term, in the weather conditions that the UK actually experiences.

So I’ve built a simulator that does. The idea is simple to state. First, describe an electricity system — so much wind, so much solar, so much nuclear and gas, so much storage of stated kinds, cables of stated sizes to named neighbours, and a pattern of demand. Then let the simulator march through forty years of recorded British weather — every half-hour from 1985 to the end of 2024, roughly seven hundred thousand steps — and at each step do what the real system would have to do: generate what the weather allows, meet the demand that the temperature implies, charge and discharge the stores, trade with the neighbours over the cables, and keep the books. When something big fails — the largest power station tripping off the grid in an instant — a second, faster part of the simulator zooms in and asks whether the system rides through the shock, second by second, or collapses into blackout.

The simulator answers questions about the physics and economics of supply meeting demand. Give it an electricity system and it will tell you whether that system keeps the lights on through forty years of real weather, how hard its storage works, what its market prices look like, whether it survives the sudden loss of its largest power station, and what it costs per unit of electricity delivered. It will answer these questions for systems that exist, systems that are planned, and systems that are merely proposed — and it will answer them identically every time, for anyone who runs it.

It makes the arithmetic of the energy transition inspectable — to replace “studies show” (whose studies?) with “here is the system, here is the weather, here is what happens; run it yourself.”

It is just as important to say what it is not. It is not a model of the electricity market, of the wires, or the gas network. It omits fuel supply chains, cyber attacks, operator error or human failure. The boundary sits where it does for a reason. Every capability inside it can be tested against recorded reality — and has been. Everything outside it would require assumptions that cannot be checked, and untestable assumptions are how energy modelling earns its poor reputation. It answers questions about the grid at a scale and with an accuracy that matters. The rule throughout: better a modest instrument whose answers can be defended than an impressive one whose answers must be believed.

grid-sim: forty years of real weather (701,280 half-hours), a wind-and-solar Britain, one 24 TWh, £200bn, £20bn/year hydrogen store. It survives — but only just, and only because it was sized for the worst week in four decades, December 1989. The case against “a few days of storage,” in one picture.

It’s the research instrument for my book. Early results are quietly impressive. It reproduces the real 2024 system to within a percent or so, and the August 2019 blackout to a few thousandths of a hertz. I’m using it now to explore real policy questions. For example, estimating the size of hydrogen stores, and what district geothermal is worth to the grid against heat pumps — the same heat, but one version arrives as demand exactly when the wind dies and the other doesn't. The gap runs to gigawatts of peak and a near-doubling of the seasonal storage the system needs. Not a rounding error.

Shortly, I’ll put the simulator on the web with some controls and you can run your own grid to explore the effects of different choices. Just don’t be surprised when you create grids that are astronomically expensive, or kill virtual people in their thousands. At least you’d be qualified to work at DESNZ.

So that’s it. It’s code name on my computer is ‘grid-sim’. There’s a prize for anyone who can think of something better. I’ll share more of its capabilities in the chat as it develops, and you can help me shape it into something useful for you too.

Richard

The Energy Trap: Why the Renewable Energy Transition Can’t Work — And What Can is published by Swift Press in September.