Estimating production loss due to icing on wind turbines

Abstract

Icing on wind turbines are known to lower their performance, but the exact relationship between wind, ice and power production is not known. In this thesis power loss due to icing on turbine blades at Aapua wind park is investigated for the winter season 2009-2010. The total loss is found to be 30%, whereof 25% is concluded being due to icing.

Three different methods are presented to estimate a power output model P(V,I), based on empirical data of wind speed, ice load and power production from a wind turbine. The models estimate power output from wind speed and ice load observations. Their performance are compared using correlation and root mean square error (RMSE), and the kriging method, using a weighted mean to calculate power output, is found best.

A comparison of the proposed kriging model to an existing model show a 10% increase in performance for the kriging method. Testing shows that kriging works well for low wind speeds and low ice loads, but tends to overestimate production during high ice loads. For the season as a whole, the modelled power output underestimates the power production with 1%.

Time series of modelled, measured and expected power output, together with ice load and temperature measurements, are investigated. Results show that sublimation, shedding, melting and accretion processes on the turbine wings, are not fully captured by the model.

Mismatch between ice loads on measuring equipment compared to turbine blades, together with insufficient number of observation data, are found to be the main reasons for inaccuracy in the model.