Icing losses

In cold climates, supercooled droplets freeze on contact with a blade and build along the leading edge. Even a thin layer changes the airfoil enough to cut output measurably, and heavier accretion causes rotor imbalance, higher loads, and a risk of ice being thrown, which is why most cold-climate sites shut down on detected icing rather than run through it.

The losses are concentrated and therefore large where they occur. Cold-climate sites can lose a substantial share of winter production to icing, and the loss lands precisely in the season with the strongest wind resource, so its effect on annual energy production is worse than the hours alone suggest.

Icing is also awkward in a performance record. An iced turbine produces far below its expectation while the wind looks perfectly normal, so if icing periods are not identified and handled, they contaminate the baseline in exactly the way curtailment does. Every serious analysis flags these periods before fitting anything.

Detection itself is usually inferential rather than direct. Few turbines carry a dedicated ice sensor, so icing is inferred from temperature, humidity, and a production shortfall that appears and disappears with the weather - the same class of signal a normal behavior model uses, applied to a condition rather than to a component.