Wind rose
How often the wind blows from each of 16 directions, split by speed at hub height.
A 700 MW onshore wind farm 50 km north-west of Yanbu, from six years of real hourly wind to the energy it sends to the grid, live today.
The farm as it would look right now.
Watch the video guideHow often the wind blows from each of 16 directions, split by speed at hub height.
Average wind at hub height for each hour of the day (local time). Arrows show where the wind comes from at that hour: the afternoon sea breeze swings it to the west.
Probability density at hub height: percent of hours per 1 m/s of speed, so the bars and the fitted Weibull curve share one scale.
Mean wind at hub height for each month. The dashed steps are the yearly averages, showing how much one year can differ from the next.
Power law v(z) = v100·(z/100)α, fitted to the 10 m and 100 m data at each hour
At night the desert surface cools and the air near the ground becomes stable, so the wind higher up decouples and α rises. In the afternoon the air is well mixed and the profile is almost flat. The blades of a 200 m rotor on a 115 m tower sweep from 15 m to 215 m, so they meet very different winds at the top and the bottom.
A 7.7 MW, 200 m rotor machine for hot, dusty sites. The attached datasheet covers the WD200 family at 50 Hz (Saudi Arabia's grid runs at 60 Hz, so Yanbu needs the 60 Hz variant); the 7.7 MW rating, 200 m rotor and 115 m hub come from the Yanbu project announcements, and the rest follows the WD200-7500.
Drag the wind speed. The rotor turns at its modelled speed; above rated wind the blades pitch, so they look thinner from the front.
Modelled from rotor size and rating: Cp ≈ 0.45 below rated, rated power reached at 12.3 m/s, cut-out 25 m/s. Hot air is thinner, so at Yanbu the curve shifts right.
From the attached Windey technical parameters (page 17). The highlighted column is the project turbine.
Each turbine leaves a slower, wider wake behind it. The colours show how much power each turbine makes right now; hover one to read it. Play a typical day or a real week from 2025 and watch the farm respond.
Rows face the prevailing north-north-west wind. Spacing is in rotor diameters (D = 200 m). Wider spacing cuts wake losses but needs more land and cable.
Wind blowing along a row loses the most. Bars are the share of energy lost; the line shows how often each direction blows.
A typical year: each calendar month averaged over all its years of hourly wind, then added up. Each red step is a loss.
P90 is the energy exceeded in 9 years out of 10: P90 = P50 × (1 − 1.282 σ). The default tariff is the SAR 0.064/kWh reported for the Yanbu PPA.
Net energy for each calendar month, averaged over all years. Summer is the windy season here.
The same farm in each data year. 2026 covers January to October only and is shown at its partial total; it was an unusually windy year in the data.
Expected output at each hour, from the Weibull spread around that hour's mean wind, after wakes and losses. Output peaks in the afternoon and evening, which lines up with Saudi summer demand.
The farm as if it were running today. Its output follows the live wind at the site: the latest forecast model run, scaled to the wind measured at Yanbu airport. The forecast gives three planning scenarios for the grid operator, and the accuracy panel scores the 15-minute and 1-hour forecasts against the measurements.
A camera at the south-west edge of the farm, looking along the rows. Every rotor faces the live wind and turns at the speed its own wind gives it; the sky follows the real sun over Yanbu.
Forecasts issued now, and how forecasts made 15 minutes and 1 hour ahead have scored against the measured wind over the last 48 hours.
Left of "now": best estimate of what the farm produced, with dots where the airport measured the wind. Right: the forecast. P50 is the expected output; the farm beats P90 nine times out of ten and P95 nineteen times out of twenty.
What the operator can count on. Commit generation against P50, hold reserves for the gap down to P90, and treat P95 as the output the farm will almost always beat.
A 5.5-minute tour of every tab: what you can change, the inputs, and what the model calculates from them, the outputs. Captions are on by default. The narration is a synthetic voice.
Welcome to the Yanbu Wind Farm Simulator, a teaching tool for a seven hundred megawatt wind farm, about fifty kilometres north-west of Yanbu. The banner shows the farm in the desert. It is drawn from the real layout, the rotors turn in the wind, and the light follows the real sun over Yanbu. Every tab works the same way. Inputs are what you choose, such as the hub height or the number of turbines. Outputs are what the model calculates from them, from the wind at the hub to the energy sent to the grid. In this guide, inputs are outlined in teal, and outputs in amber. Let's go through the tabs one at a time.
First, the four controls at the top of every tab. Hub height moves the rotor up or down. A taller tower reaches faster wind, so the energy rises. The ERA5 wind correction scales the historical wind data, to test how much the answer depends on it. Turbines sets the size of the farm, and air density switches between the hot, thin air of the site and standard air. On the right is the headline output: the net energy in a typical year, and the capacity factor.
The Wind resource tab describes the wind itself, from almost six years of hourly data. Its inputs are the period buttons: all years, a single year, or one calendar month. The outputs start with six summary tiles: the mean wind at hub height, the Weibull shape and scale, the power density, the prevailing direction, the wind shear, and the air temperature and density. The wind rose shows where the wind comes from. Switch it to energy, and the north-west stands out: it brings most of the energy. The daily profile shows the sea breeze. The wind is weakest in the morning and strongest in the late afternoon. Further down, the histogram shows how often each wind speed occurs, and the shear chart shows how the wind grows with height.
The Turbine tab shows one Windey turbine, rated at seven point seven megawatts, with a two hundred metre rotor. The input is the wind speed slider. As the wind rises, the rotor turns faster, the blades pitch, and every readout follows. The outputs are the power curve, at standard and at site air density, the power and thrust coefficients, and the datasheet values. Above about twelve and a half metres per second, the power stays at seven point seven megawatts. The blades pitch to spill the extra wind.
The Wind farm tab puts ninety-one turbines together. One set of inputs chooses the wind: a manual speed and direction, a typical day for any month, or a real week from 2025. The other set is the layout: the number of rows, their direction, the spacing, the wake decay, and staggered or aligned rows. The output is the map. Each turbine leaves a slower, wider wake behind it, and its colour shows how much power it makes. The timeline replays the day hour by hour, and further down, the chart gives the wake loss for each wind direction.
The Energy yield tab turns the wind into energy over a typical year. Its inputs are the five losses: availability, electrical, environmental, performance and curtailment, plus the uncertainty and the electricity tariff. The outputs are the net energy, the capacity factor, the full load hours, the P90 energy and the revenue. The waterfall chart starts from the gross energy and takes away each loss in turn, down to the net energy delivered to the grid. Below it, the monthly, yearly and hourly charts show when the energy comes. June to September give about half of the year.
The Live and forecast tab runs the farm on today's wind. Its inputs are live data rather than sliders: the latest forecast model run at the site, the wind measured at Yanbu airport, and a fifty-one member ensemble forecast. In this recording, they are replaced by example data. The outputs start with the live tiles and a camera view of the farm. Each rotor faces the live wind, and turns at the speed of its own wind. Next are the forecasts for the next fifteen minutes and the next hour, scored against the measured wind. The chart covers the next seventy-two hours in three scenarios. P50 is the expected output. The farm beats P90 nine times out of ten, and P95 nineteen times out of twenty. As each hour passes, the forecast check compares the forecast with the actual output, and the Excel button saves the whole record. Finally, the scenario table tells the grid operator how much energy to count on for the next day.
The Method and data tab lists the data sources, the equations and the student exercises. That completes the tour. Change an input, watch the outputs respond, and you will see how a wind farm turns the Yanbu wind into electricity.
Hourly ERA5 reanalysis from the Copernicus Climate Change Service, retrieved through the Open-Meteo archive API for 24.41° N, 37.71° E, about 50 km north-west of Yanbu city, on land at 108 m elevation (ERA5 grid cell 24.50° N, 37.75° E). Period 1 January 2021 to 31 October 2026, local time (UTC+3), 51,120 hours with no gaps. Variables: wind speed and direction at 100 m, wind speed at 10 m, air temperature at 2 m and surface pressure.
The simulator stores 5.8-year statistics computed from every hour: speed histograms in 0.5 m/s bins, a 16-sector × 1 m/s direction table, hour-of-day averages, shear and air density, for all years, each year and each calendar month. It also stores four real hourly weeks from 2025 for the replay.
Caution. ERA5 has a 0.25° (about 28 km) grid. It smooths the coastal mountains and valleys and often underestimates wind at well-exposed sites. A bankable study would correlate a year or more of on-site met-mast or lidar data with the long-term reanalysis. Use the correction slider to explore that uncertainty.
WD200 family data come from the attached Windey technical parameters. The project's 7.7 MW rating, 200 m rotor, 115 m hub height, the 700 MW capacity and the location about 50 km north-west of Yanbu are from public project announcements; press reports mention 87 turbines, while 91 units are needed for 700 MW at 7.7 MW each. The power curve is modelled (Cp = 0.45 with a smooth approach to rated), not the manufacturer's certified curve, and the thrust curve is generic.