PowerVeritas
Insights / Field note · Wind

Two icing signatures on a power curve, only one is genuine

A power curve can show two patterns that both look like icing. The time series context reveals that only one of them actually is.

← All insights

This power curve contains two potential icing issues, but only one is genuine icing.

Turbine power against wind speed coloured by ambient temperature, with the design power curve in red. A green box highlights scatter to the left and above the curve, a blue box highlights scatter to the right and below the curve, both occurring near zero degrees.
Power against wind speed for one turbine, coloured by ambient temperature, with the design curve in red. Two clusters of scatter sit off the curve, highlighted by the blue and green boxes.

The scatter to the right of the power curve, in the blue box, occurs around 0°C and is caused by blade icing. Ice on the blades degrades their aerodynamic performance and can cause sections to stall, leading to underperformance and often a low torque icing alarm that shuts the turbine down.

The scatter to the left of the power curve, in the green box, also occurs mostly around 0°C. At first glance this looks like anemometer icing, on a cup rather than a sonic sensor, where ice on the cups slows the anemometer down, causing it to under-read and making the turbine appear to over-produce.

A closer look at the temperature trend at the onset of each issue tells the real story. The blade icing scatter starts after temperatures drop below 0°C and resolves once they rise back above. The other issue starts when ambient temperature is around 9°C, which is well above icing temperatures, even though temperatures do then drop to near 0°C during the fault period. The actual issue is a faulty primary anemometer.

Time series of wind speed and ambient temperature for two neighbouring turbines, with vertical markers for onset and resolution of the issue. The wind speed traces diverge at the onset while ambient temperature is around 9 degrees.
The temperature trend for the other issue, the green-box scatter. Wind speed and ambient temperature for the faulty turbine 1 and neighbouring turbine 2, with markers for onset and resolution. The wind speed traces diverge at onset while temperature is around 9°C, well above icing, pointing to a sensor fault rather than icing.

The lesson

Looking at the wider time series context, not just the power curve in isolation, is essential for getting the diagnosis right.

Charts and analysis by PowerVeritas. Where open datasets are used, sources are credited on the attributions page.