Wind potential at Whittlesea, South Africa: Weibull from mean speed and power
A year of hourly winds in the Eastern Cape gives a mean of 3.88 m/s and 67 W/m² at 10 m. We back out the Weibull curve, correct for the thin air at 1,100 m and size a small turbine with fibub's wind energy tool.
Before anyone puts up a wind turbine, they need to know how often each wind speed blows. Engineers describe that with a Weibull distribution, a curve set by two numbers: a shape k (how spread out the speeds are) and a scale c (roughly, a typical speed). A 2025 study in Processes compares six ways of finding k and c from a year of hourly winds at Whittlesea in South Africa’s Eastern Cape, a rural area looking at small turbines for local power. This article shows how the wind energy yield tool turns the paper’s two headline numbers back into a Weibull curve, fixes the air density for a high site, and estimates what a small turbine would make. For the same site’s sunshine, the solar PV yield tool does the solar side.
The paper in brief
The authors took one year (2022) of hourly mean wind speeds measured 10 m above the ground near Whittlesea, at about 1,100 m above sea level. They fitted the Weibull curve six ways: five classic numerical methods (such as the method of moments and maximum likelihood) and a newer search method, the whale optimisation algorithm, then compared how well each matched the data. The site’s mean wind speed is 3.88 m/s, rising to 4.87 m/s in winter, and its wind power density (the wind’s power through each square metre facing it) is 67.29 W/m², worked out with the standard sea-level air density of 1.225 kg/m³. That is a low-wind site, suited to small turbines rather than wind farms.
Checked with the tools
We started from only the paper’s mean speed and power density, plus its site elevation. Everything here is a recomputation with the tool’s maths and our stated assumptions, not a fit to the paper’s hourly data.
Step 1: back to the Weibull curve. For a Weibull curve, the ratio of the power density to ½ρv̄³ (ρ is the air density, v̄ the mean speed), called the energy pattern factor, depends only on k. The paper’s numbers give 1.881, which fixes k = 2.03; the mean then gives c = 4.38 m/s. Fed back into the tool’s formulas, these return the paper’s 3.88 m/s and 67.29 W/m² exactly (Figure 1). A k near 2 is typical of inland sites.
Step 2: thin air. At 1,100 m and an assumed 15 °C, the tool’s standard-atmosphere maths gives an air density of 1.073 kg/m³, about 12% below the 1.225 the paper used. Wind power scales with density, so the site’s real power density at 10 m is about 59 W/m², not 67 (Figure 2). Both values are below 100 W/m², the bottom wind class on the classic scale.
Step 3: higher up, and a small turbine. Wind is faster away from the ground. With the usual 1/7 power law, the mean rises to 4.54 m/s at 30 m and the power density to 94 W/m² (site air). A made-up 10 kW turbine with an 8 m rotor would make, before losses:
| Hub height | Mean speed | Power density (site air) | Energy a year | Capacity factor |
|---|---|---|---|---|
| 10 m | 3.88 m/s | 59 W/m² | 9,979 kWh | 11.4% |
| 20 m | 4.28 m/s | 79 W/m² | 13,244 kWh | 15.1% |
| 30 m | 4.54 m/s | 94 W/m² | 15,486 kWh | 17.7% |
| 50 m | 4.88 m/s | 118 W/m² | 18,649 kWh | 21.3% |
Raising the hub from 10 to 30 m adds about 55% more energy: at a low-wind site, height matters more than almost anything else.
How we made the graphs
Both charts and the table come from scripts/blog-figures/whittlesea-wind-weibull-power-density.mjs in fibub’s public code. It calls gamma, cFromMean, powerDensity, weibullPdf, shear, genericCurve and aep from src/lib/sci/eng-wind.ts, and atmosphere from src/lib/sci/eng-atmos.ts: the maths behind the wind energy yield tool. The mean speed, power density, measuring height and elevation come from the paper. The temperature, the shear exponent and the turbine are our assumptions, and the turbine’s yearly energy is a model estimate, not a result from the paper. The paper’s six fitted k and c pairs are in its tables; ours is a single back-calculation that matches its two summary numbers.
How we test the maths
On every change to fibub, tests/sci-tools/wind-energy-yield.mjs (run by tests/sci.mjs) checks the tool against SciPy:
- The maximum-likelihood Weibull fit of a fixed example year must give k 2.10163 and c 7.17262 m/s, matching an exact solve with SciPy’s brentq.
- The annual energy of a generic 2 MW turbine on that wind must be 10,308,664 kWh at 1.225 kg/m³ and 9,925,074 kWh at 1.1 kg/m³, matching SciPy’s quad integration.
- New with this article: k 2.0312 and c 4.3792 m/s must give back a mean of 3.88 m/s and a power density of 67.29 W/m².
Check it yourself
On the wind energy yield tool:
- Choose Weibull k and c in Wind from. Type 2.03 in Shape k and 4.38 in Scale c (m/s).
- Type 10 in Measured at height (m) and 10 in Hub height (m). Leave Site elevation (m) on 0 and Mean air temperature (°C) on 15.
- It reads a Mean speed at hub of about 3.88 m/s and a Wind power density of about 67 W/m²: the paper’s numbers.
- Now type 1100 in Site elevation (m): Air density drops to about 1.073 kg/m³ and the power density to about 59 W/m².
- Type 30 in Hub height (m) (leave Shear exponent α at 0.143). With Power curve on A generic made-up curve, type 10 in Rated power (kW), 8 in Rotor diameter (m), 0.4 in Power coefficient Cp, 3 in Cut-in speed (m/s) and 25 in Cut-out speed (m/s); set Availability (%) to 100 and Other losses (%) to 0. It reads about 15,500 kWh a year, capacity factor 17.7 %.
What researchers can use it for
- Rebuild a Weibull curve from the mean speed and power density a paper reports, to compare sites that were fitted with different methods.
- Correct for air density at high or hot sites, a common gap in published resource studies that use 1.225 kg/m³.
- Fit your own data: paste a year of hourly speeds and the tool finds k and c by maximum likelihood, the method the paper’s comparison includes.
- Screen hub heights and turbines for small-scale projects before a full study, with the maker’s power curve pasted in.
- Teach why a mean speed alone is not enough: two sites with the same mean but different k have different power.
The fine print
- One year of data at one station; wind varies from year to year by several per cent, and the station is some distance from the village.
- The 1/7 shear law is a rule of thumb; real shear depends on the ground, trees and buildings, and on the time of day.
- The turbine is made up, and the energy is before availability and other losses, which usually take 10 to 20% off.
- Our k and c match the paper’s summary numbers but are not any of its six fitted pairs.
Sources
- Shambira, N., Luvatsha, L. and Mukumba, P. (2025), Comparative Analysis of Five Numerical Methods and the Whale Optimization Algorithm for Wind Potential Assessment: A Case Study in Whittlesea, Eastern Cape, South Africa, Processes 13(5), 1344, doi:10.3390/pr13051344. Licence CC BY, checked on the journal page, 2026-11-16. Values used here: hourly wind speeds for January to December 2022 at 10 m above ground (South African Weather Service, Queenstown station); site elevation 1,100 m; mean speed 3.88 m/s (winter mean 4.87 m/s); wind power density 67.29 W/m² with an air density of 1.225 kg/m³.
- Air density from the International Standard Atmosphere (ISO 2533:1975), as used by the Wind energy yield tool.
- Wind power classes (class 1 below 100 W/m² at 10 m): Elliott, D. L. et al. (1986), Wind Energy Resource Atlas of the United States, Pacific Northwest Laboratory, DOE/CH 10093-4.
- Our assumptions, not from the paper: mean air temperature 15 °C; power-law shear exponent 0.143 (1/7, open land); a made-up 10 kW turbine with an 8 m rotor, power coefficient 0.4, cut-in 3 m/s and cut-out 25 m/s.
If you are an author or publisher and want a credit or a passage changed or removed, tell us and we will act quickly.