Wind Turbines & the Betz Limit

Why a turbine must let the wind keep going

Junior level — plain language, no maths

Wind is just air with somewhere to be. It carries energy because it is moving, and a turbine's job is to steal some of that motion on the way past. What surprises people is how much energy is in there, and how brutally it depends on speed. Double the wind and you do not get twice the power - you get eight times it. Two things are doubled at once: how fast each parcel of air is moving, and how many parcels arrive each second. Multiply that through and the power goes as the cube of the speed.

Which is why the wind map matters more than the machine. A site with wind 20% stronger delivers about 70% more energy from exactly the same turbine, and why developers argue so fiercely about hub height: climb 50 metres up and you leave the friction of the ground behind, where the wind is faster and smoother.

Now the beautiful part. You might think the perfect turbine takes all the energy out of the wind. It cannot - and not because we are bad at building them. If the blades really did take everything, the air behind them would be at a dead stop. Stopped air does not get out of the way, and air that does not get out of the way blocks everything arriving behind it. Take it all and nothing more comes through. Take nothing, and air sails past untouched. Somewhere between those two failures sits the best possible compromise.

Albert Betz worked out where in 1919. Slow the wind to two thirds of its original speed as it passes the blades, let it leave at one third, and you extract 59.3% of the energy in the air - the most any open rotor can ever take. Real turbines manage 45 to 50%, which is remarkably close to a theoretical ceiling. And they only run within a window: below about 3 m/s there is not enough to bother turning, above 25 m/s they turn their blades edge-on to the gale and shut down, because surviving the storm is worth more than the electricity in it.

The cube law, the actuator disc and where 16/27 comes from

Student level — the core equations

Start with kinetic energy. A parcel of air of mass \(m\) moving at \(v\) carries \(\tfrac{1}{2}mv^2\). The mass arriving each second through a disc of area \(A\) is \(\dot m = \rho A v\). Multiply: the power crossing that disc is \(P = \tfrac{1}{2}\rho A v^3\). The cube is not a modelling choice, it is that \(v\) appears once in the energy per unit mass and once again in how much mass shows up.

Now treat the rotor as an actuator disc - a permeable surface that removes momentum without caring how many blades do it. Define the axial induction factor \(a\) by saying the wind has already slowed to \(v(1-a)\) by the time it reaches the disc. Momentum conservation then forces the far wake to \(v(1-2a)\): the disc sits exactly halfway through the total slowdown. The thrust is \(T = \dot m\,\Delta v = \rho A v(1-a)\cdot 2av\), and the power extracted is that thrust times the speed at the disc.

Put the pieces together and \(P = 2\rho A v^3 a(1-a)^2\), which as a fraction of the power in the wind gives the power coefficient \(C_p = 4a(1-a)^2\). This function is zero at both ends and humped in the middle, so differentiate: \(dC_p/da = 4(1-a)(1-3a)\), which vanishes at \(a = 1/3\). Substituting back gives \(C_p = 16/27 = 0.593\). That is Betz's limit, and notice what it did not require: no blade shape, no aerodynamics, no material. Only mass and momentum.

Real machines run at \(C_p \approx 0.45\text{–}0.50\), losing the rest to wake rotation, finite blade count and drag. They also live inside a control envelope. Below cut-in there is not enough torque to overcome losses. Between cut-in and rated, the controller varies rotor speed to hold the tip speed ratio \(\lambda = \omega R/v\) near its optimum of about 7, keeping \(C_p\) at its peak. Above rated wind the generator is at its limit, so the blades pitch out of the wind and deliberately throw energy away to hold power flat. Above cut-out, everything feathers and stops.

Momentum theory, blade elements, and the square-cube fight that decides turbine size

Scholar level — full mathematical depth

01The actuator disc, honestly stated

One-dimensional momentum theory replaces the rotor with a permeable disc across which pressure drops discontinuously while axial velocity stays continuous. Applying Bernoulli separately upstream and downstream of the disc - you cannot apply it across, because energy is being removed - gives \(\Delta p = \tfrac{1}{2}\rho(v_1^2-v_2^2)\), so thrust \(T = A\Delta p\). Equating that to the momentum flux \(\dot m (v_1-v_2)\) with \(\dot m = \rho A v_d\) forces \(v_d = \tfrac{1}{2}(v_1+v_2)\): the disc velocity is the arithmetic mean of the far-field speeds. Everything else follows, including \(P = 2\rho A v_1^3 a(1-a)^2\) and the streamtube expansion \(A_1 v_1 = A_d v_d = A_2 v_2\) that makes the flow visibly fatten as it passes.

02What the limit actually forbids

Maximising \(C_p = 4a(1-a)^2\) gives \(16/27\) at \(a = 1/3\), and the momentum derivation itself breaks down beyond \(a \approx 0.4\), where the wake becomes turbulent and empirical corrections take over. Three caveats are worth stating precisely. Wake rotation - the torque reaction spins the wake - costs a further few per cent, more at low tip speed ratio, which is the Glauert correction. Finite blade count leaks flow around the tips; Prandtl's tip-loss factor accounts for it. And the limit is defined against the rotor disc area: a diffuser-augmented turbine can exceed \(16/27\) of its rotor area precisely because it draws in a streamtube wider than the rotor, which is not a violation but a change of the denominator.

03From a disc to actual blades

Blade element momentum theory closes the gap. Divide the blade into radial elements, each an aerofoil seeing a relative wind that combines the axial flow \(v(1-a)\) with the rotational flow \(\omega r(1+a')\), where \(a'\) is the tangential induction factor. Each element produces lift and drag from its local angle of attack, integrate for thrust and torque, then iterate until the element loads and the momentum balance agree. The design payoff is that lift dominates: modern rotors are lift machines running at tip speed ratios near 7, where the blade tip travels several times faster than the wind, which is why three slender blades beat a sail-like rotor with dozens of them.

04Control, in four regions

Below cut-in, aerodynamic torque cannot overcome losses. In region 2 the controller applies generator torque \(Q = K\omega^2\) with \(K = \tfrac{1}{2}\rho\pi R^5 C_{p,\max}/\lambda_{\text{opt}}^3\), which - remarkably - holds \(\lambda\) at its optimum without ever measuring the wind. In region 2.5 the rotor hits its speed limit and \(\lambda\) drifts. Above rated wind, region 3, the machine already produces all the power the generator and cables can take, so collective pitch turns the blades into the wind to shed lift and hold \(P\) flat - deliberately spilling energy for the rest of the curve. Above cut-out, typically 25 m/s, the rotor feathers and idles, because the design loads that matter are structural, not electrical.

05Why turbines keep getting bigger

Energy capture scales as \(D^2\) while a naive scaling of blade mass goes as \(D^3\) - the square-cube law, which ought to have stopped growth decades ago. It has not, because the cube is not obeyed: carbon spar caps, slender high-aspect-ratio blades, segmented moulds and direct-drive generators have kept the mass exponent nearer 2.3. Height helps twice over, since the shear profile \(v(z) = v_r(z/z_r)^\alpha\) with \(\alpha \approx 0.14\) over open ground means a taller hub sits in faster, less turbulent air, and cubing that gain magnifies it. The binding constraints today are logistics and blade-root loads, not aerodynamics.

06From one machine to a wind farm

Annual energy is not the power curve at the mean wind but the power curve integrated against the wind's distribution, usually Weibull with shape \(k \approx 2\): \(E = 8760\int_0^\infty P(v)f(v)\,dv\). Because \(P \propto v^3\) below rated, variance is worth money. Then put the machines together and they interfere: each wake carries a velocity deficit that recovers slowly, and array losses of 5-15% are typical, which is why spacing runs 7-10 diameters downwind. The simple Jensen model, \(v_w = v\left[1-2a/(1+2kx/D)^2\right]\), still underpins layout optimisation. What binds at last is the grid: an asynchronous, inverter-coupled fleet contributes no natural inertia, so the same machines that make the cheap electrons also make frequency control harder - an accounting problem the whole system now has to solve.