What capacity factor measures, and what it does not
Capacity factor is a ratio of energy to energy. Take what a plant actually delivered over a period, divide by what it would have delivered running at nameplate every hour of that period, and express the result as a percentage. A 3 MW turbine that produces 9,500 MWh in a year has a capacity factor of 9,500 ÷ (3 × 8,760) = 36.1%.
The number is often misread as an efficiency, and it is not one. A wind turbine at 36% capacity factor is not wasting 64% of anything. It is a fuel-availability measure: the wind blows at the speed the machine needs for only part of the year, and the rest of the time the machine is either below rated output or stopped. A thermal plant with the same capacity factor is telling you something completely different — that it is running part-time because of dispatch economics or outages.
Because capacity factor divides by nameplate, it also depends on a design choice rather than only on the resource. Fit a smaller generator behind the same rotor and the capacity factor rises even though the energy falls slightly, because the denominator fell more than the numerator did. That is the specific-power trade-off that drove the industry toward larger rotors on relatively smaller generators, and it is why capacity factor alone never settles which of two turbines is better for a site. Pair it with the power output calculation and the levelized cost of energy before drawing conclusions.
From a wind distribution to annual energy
Measured capacity factor is arithmetic. Modelling one before the turbine exists is the interesting problem, and it has two ingredients: how often each wind speed occurs, and how much power the machine makes at that speed.
The wind distribution. Hourly wind speeds at a site are well described by a Weibull distribution with shape factor k and scale factor c. The shape factor controls how tightly speeds cluster: k = 2 gives the Rayleigh distribution, the usual default for open temperate sites, while steadier trade-wind sites run nearer 3 and gusty inland or mountain sites nearer 1.5. You rarely know c directly, but you usually know the mean, and the two are linked by c = v̄ / Γ(1 + 1/k). At k = 2 that gamma function is Γ(1.5) = 0.886227, so c = 1.1284 v̄.
The power curve. Below cut-in the turbine makes nothing. Between cut-in and rated speed, aerodynamic power rises with the cube of speed, so the model ramps as (v³ − v_in³)/(v_rated³ − v_in³) of nameplate. From rated speed to cut-out the controller holds output flat at nameplate. Above cut-out the machine stops. That four-segment idealisation reproduces the shape of a manufacturer's measured curve closely enough for screening work.
The integral. Multiply the two functions and integrate across all wind speeds. The result is the mean power the turbine produces over a long period, and dividing by nameplate gives the gross capacity factor directly. Multiply by the hours in the year for gross annual energy production, then deduct the loss stack for net.
One consequence worth internalising: because the mean of v³ exceeds the cube of the mean, a site's usable resource is always higher than its mean speed alone suggests. For a Rayleigh distribution the mean cube is exactly (6/π) times the cube of the mean — about 1.91 times — and that identity is the basis of one of this calculator's test cases.
Worked example: both routes on one 3 MW turbine
Route one, from meter readings. A 3 MW turbine delivered 9,500 MWh over a full year.
- Maximum possible energy. 3 MW × 8,760 h = 26,280 MWh.
- Capacity factor. 9,500 ÷ 26,280 = 0.36149 = 36.15%.
- Equivalent full-load hours. 9,500 ÷ 3 = 3,167 h, which is also 0.36149 × 8,760.
Route two, from the wind. The same machine has a cut-in of 3 m/s, rated speed 12 m/s and cut-out 25 m/s. The site's long-term mean at hub height is 8 m/s with a Rayleigh distribution.
- Scale factor. c = 8 ÷ 0.886227 = 9.027 m/s.
- Integrate. Weighting the four-segment power curve by the Rayleigh density gives a mean output of 1.133 MW.
- Gross capacity factor. 1.133 ÷ 3 = 37.77%.
- Gross annual energy. 0.3777 × 3 × 8,760 = 9,926 MWh.
- Apply 15% losses. Net capacity factor = 37.77 × 0.85 = 32.10%; net energy = 8,437 MWh.
The two routes disagree — 36.15% measured against 32.10% modelled. That is a normal and informative result. It says either the year was windier than the long-term mean, or the 15% loss assumption is too heavy for this site, or the mean speed used in the model is conservative. A single year's generation is a noisy estimate of a long-term capacity factor, which is why bankable assessments run multiple years of on-site measurement correlated against long-term reference data.
What counts as a good capacity factor
Judge a wind capacity factor against three things, in this order: the site's mean wind speed, the turbine's specific power, and the loss assumption. Without all three, a bare percentage is not comparable to anything.
The reference table below gives the arithmetic version of the first point. Holding the turbine and losses fixed, the modelled net capacity factor climbs from about 5% at a 4 m/s mean to about 51% at 12 m/s. That range is not a spread of turbine quality; it is entirely resource. It is also why the industry's usual screening threshold for a viable site sits well above the mean speed most inland low-rise locations offer.
Watch the gross-to-net step carefully, because it is where optimistic assessments hide. Gross energy is what the power curve and the wind distribution alone predict. Net is what reaches the meter, after wake interference between machines, turbine availability, electrical collection and transformer losses, blade degradation and soiling, curtailment for grid or environmental constraints, and any icing or high-temperature derating. Each is small; together they are not.
Finally, note what a rising capacity factor does not automatically mean. Because it is energy divided by nameplate, a machine with a lower rated power behind the same rotor shows a higher capacity factor while producing slightly less energy. If you are comparing two turbine models for the same site, compare annual energy production and cost, not capacity factor. If you are comparing two sites for the same turbine, capacity factor is exactly the right measure.
Modelled capacity factor against mean wind speed
| Mean wind speed (m/s) | Gross CF (%) | Net CF (%) | Net annual energy (MWh) |
|---|---|---|---|
| 4 | 5.94 | 5.05 | 1,327 |
| 5 | 12.29 | 10.45 | 2,745 |
| 6 | 20.45 | 17.38 | 4,568 |
| 7 | 29.27 | 24.88 | 6,537 |
| 8 | 37.77 | 32.10 | 8,437 |
| 9 | 45.35 | 38.54 | 10,129 |
| 10 | 51.65 | 43.90 | 11,538 |
| 11 | 56.52 | 48.04 | 12,626 |
| 12 | 59.95 | 50.95 | 13,391 |
Rows are produced by the same integration the calculator performs. Note how the gain per extra metre per second shrinks at the top: the machine is already at rated output for much of the time, so extra wind is spilled rather than captured.
Assumptions and traps
- Mixing gross and net. A modelled gross figure compared against a measured net figure will always flatter the model. Compare like with like, and state which you are quoting.
- Using a 10 m mast speed as the hub-height mean. Shear means hub-height wind is faster, and capacity factor responds steeply to that difference. Correct for height before entering a mean speed.
- Assuming Rayleigh without checking. k = 2 is a convention, not a measurement. Fitting the actual shape factor to site data changes the answer, particularly for sites whose wind is either unusually steady or unusually gusty.
- Reading capacity factor as efficiency. It is a measure of how much of the time the resource lets the machine run near nameplate, not of how well the machine converts what it receives. The power coefficient is the efficiency measure.
- Comparing across technologies without saying so. A solar array, a wind farm and a nuclear plant with the same capacity factor have almost nothing in common. The number is only comparable within a technology, or when what you care about is literally the ratio of energy to nameplate.
- Judging a project on one year. Interannual wind variability of several percent in energy terms is normal. A single low year is not evidence of underperformance, and a single high year is not evidence of a good site.
- Forgetting the denominator can change. If turbines are added, uprated or derated mid-period, the rated capacity in the denominator is not a single number, and the simple ratio stops being meaningful.
The idealised power curve is a screening tool
The cubic ramp between cut-in and rated speed used here matches the physics of the wind power equation and the general shape of real machines, but manufacturers publish measured curves that differ in detail — particularly near cut-in, where control strategy dominates, and just below rated, where the pitch system starts to intervene. For a feasibility screen the difference is small. For an energy assessment that anyone will lend against, use the certified power curve, measured to the applicable IEC test standard, and integrate against a site-specific distribution rather than a single mean.
Key terms
- Capacity factor
- Energy delivered divided by the energy that would be delivered at rated power for every hour of the period, expressed as a percentage.
- Equivalent full-load hours
- The number of hours at nameplate output that would produce the same energy. It equals capacity factor times the hours in the period, and is the same information in a different unit.
- Weibull shape factor k
- The parameter controlling how tightly wind speeds cluster around the mean. Lower k means a broader spread with more calm and more strong-wind hours; k = 2 is the Rayleigh special case.
- Wake loss
- The energy a turbine loses because it stands in the slowed, more turbulent air behind another turbine. It depends on spacing, layout and wind rose, and is one of the largest items in a wind farm's loss stack.
- Specific power
- Rated generator power divided by rotor swept area, in W/m². Lower specific power means a bigger rotor for the same generator, which raises capacity factor.
Where capacity factor fits in a project assessment
A wind energy assessment runs in a fixed order, and capacity factor appears near the end rather than the beginning. First comes measurement: an on-site met mast or remote-sensing campaign, usually a year or more, correlated against a long-term reference series to remove the bias of an unusual measurement period. Then vertical extrapolation to hub height, then a flow model to move the measured point to each turbine location. Only then does the power curve get integrated, producing gross energy. The loss stack turns that into net energy, and net energy divided by rated capacity times 8,760 is the net capacity factor.
From there the number feeds two places. It goes into the revenue line of the financial model, where it is combined with a price per MWh to produce cash flow, and it goes into the levelized cost of energy, which is what makes the project comparable to other generation sources. For distributed and behind-the-meter machines, the value of the energy depends on the tariff rather than a market price, which brings the net metering arithmetic back into play.
The same capacity-factor definition applies unchanged to any generator. Run it on a hydro scheme sized with the micro-hydro calculator, or on a solar array whose output you have estimated with the production calculator, and the arithmetic is identical — only the physical reason for the shortfall against nameplate changes.