One of the most puzzling studies on the viability of providing global electricity from primarily wind and solar is Mark Z. Jacobson‘s and Mark A. Delucchi’s 2011 paper “Providing all global energy with wind, water, and solar power, Part I: Technologies, energy resources, quantities and areas of infrastructure, and materials,” published in Energy Policy. The paper has been cited more than 280 academic papers so far, and tends to crop up all over the place when discussing alternative energies. Let’s look at some of their assumptions. In their introduction they state that they’ll only consider
- “options … that can be scaled up as part of a global energy system without further major technological developments”
- “technologies that have low impacts on wildlife, water pollution, and land”
- “do not have significant waste-disposal or terrorism risks associated with them”
They then go on to argue that wind power meets these criteria best, followed by concentrated solar.
I’m not entirely sure how wind turbines have low impacts on wildlife and land, considering the just under 50,000 wind turbines in the US already kill entire populations of bats and hundreds of thousands of birds – while Jacobson and Delucchi recommend to build at least 540,000 more by 2030. I’m also not going to discuss why their dismissal of nuclear power is wrong – I’ll leave that to others.
I’m going to stick entirely to basic wind power facts. Jacobson and Delucchi that by 2030, total global electricity demand will be about 100,740TWh, requiring approximately 11.5TW installed capacity, approximately five times of current values. So far, so reasonable.
But it’s what follows makes very little sense, and seems completely wrong.
They then suggest that 50% of this demand (50,370TWh) can be met by 3.8 million turbines, with a capacity of 5MW (5.75TW installed capacity) taking up about 1.17% of the global land surface. But to arrive at these numbers they make some very unrealistic assumptions. For example, to generate 50,370TWh with 5.75TW installed wind capacity assumes a capacity factor of 30.26%, that is: they produce on average 30.26% of their nameplate capacity. That is highly optimistic at best, and misses some important points.
It is true that current wind power systems in the US have an annual capacity factor of about 32%. Some wind projects in the UK and Denmark have reached annual capacity factors of close to 50%. But that is not enough. Annual capacity factors don’t tell us anything about how much electricity a turbine will produce at any given time, which may range from 0% to 100%.
What matters more than average capacity factors is the ability for wind to replace conventional power sources at any given time, also known as the replacement factor: how many units of wind capacity needs to be installed to replace one unit of conventional capacity? When wind power provides only a small fraction of the total energy generating capacity – say 1% – the replacement factor does not matter very much. Even if there is no wind, it is no serious challenge to increase power production of other generating stations by 1%. However, as the share of wind increases, the ability of the overall system to absorb these fluctuations becomes diminished, and it becomes necessary to overbuilt wind capacity to ensure it generates sufficient electricity even when there is little wind. As the share of wind generating capacity increases, the replacement factor goes down – to the point that it if wind reaches requires 25 units of installed wind capacity to replace a single unit of conventional capacity – regardless of the average capacity factor.
Jacobson and Delucchi’s calculations do not appear to account for the diminishing returns on installed wind capacity and instead appear to assume that the capacity factor is all that matters. Once the replacement factor is taken into consideration, the numbers for how many 5MW turbines will be required to generate 50% of electricity in 2030 are quite different from those calculated by Jacobson and Delucchi: 28,75 Million, not 3.8 Million.
Jacobson and Delucchi further assume that the total land surface required for their 3.8 million wind farm of 5MW turbines would take 1.17% of global land area. I did not see them reference any specific wind turbine, but I will base my calculations on the 526 Hysosung 5MW offshore wind turbines, with a diameter of 139m. (In my calculations, I generally assume a minimum spacing of 6x diameter. The exact numbers depend on complex factors, but using a formula with 6x diameter (approximately 5x diameter between rows, and 7x diameter between turbines) provides a very good minimum space requirement for wind farms when compared to existing projects. ) Using a more realistic 6x turbine diameter spacing, 3.8 million turbines with a diameter of 139m would require a wind farm of at least 2,075,896 square kilometers – equivalent to 1.39% of global land area. Not 1.17%. To reach their number, the spacing is closer to 5.5x diameter. That’s extremely tight spacing – even by current standards.
Still, 1.17% of global land surface is equivalent to 12.2% of Russia, 20.8% of Canada, 105.2% of Mexico, and 379.5% of France, or 10,949% of Lake Ontario.
However, when combined with the more realistic 28.75 million turbines required to generate 50% global energy demand, and using a minimum spacing of 6x diameter, one would have to build a windfarm covering 15,705,792 square kilometer – that is 10.5% of the global land area, or 92% of Russia, 157% of Canada, 2,871% of France, or 82,836% of Lake Ontario.*
How much would it cost to build this many turbines? If we generously assume 1MW of nameplate capacity costs about $1.3M (the lower bound right now), and that ramping up production by at least an order of magnitude per year will not increase prices, building 28.75 million turbines would cost at least $187 trillion – more than 10 times current US GDP. For Jacobson and Delucchi’s calculations to be defensible, they must address the replacement factor issue. I don’t see how they can, though – and I must conclude that their entire thesis is based on a catastrophic conceptual error.
*An earlier version of this article had included a silly mistake – I had accidentally used the total surface of the earth to derive Jacobson and Delucchi’s spacing assumptions. That had two effects: I thought they had made a surprisingly realistic assumption on spacing (about 10x turbine diameter vs my 6x), while also creating a much greater space requirement for the total number of turbines that are required to meet 50% of global demand by 2030. Unfortunately, once I plugged in the correct number for the global land surface it appears their assumptions on turbine spacing are too tight by almost 10% even by current standards – and therefore underestimate land requirement for their 3.8 million turbines by a non-trivial 15.8%. Since new research seems to suggest turbines should be placed apart an even greater distance – possibly up to 8x turbine diameter – space requirements of a realistic wind farm to supply this much power will be even greater. However, I will stick to the 6x diameter standard to avoid accusations of being unfairly biased against wind.