Wind farms are nowadays more often constructed in alpine terrain than earlier due to theprofitable wind resource as well as, often, less conflicting interests than in more denselypopulated areas. The cold climate poses a relatively new challenge to the wind power industrysince icing of the wind turbine blades and sensors may induce losses in production, increasethe wear and tear of the components, leading to a shortening of structural life time as well as itdecreases the availability and hence reducing the economical profitability for the owner.This study focuses on modelling of ice accretion on a vertically mounted cylinder,dimensioned to correspond to an IceMonitor, and comparing the results with measured iceload on a similar instrument during the winter of 2009/2010. The modelling is carried outwith both a statistical approach using multiple linear regression and a physical approach usingmodel for ice accretion. Ice load was also modelled for the period 1989-2009 using the ERAinterimre-analysis data set in order to compare the winter 09/10 with a longer referenceseries. Modelled ice loads for four winters between 2005 and 2009 were compared withproduction data to investigate a possible connection between ice load and production losses.The results showed that the statistical approach was unable in its current form toreproduce and predict measured ice loads and the method was deemed unsuitable. Thephysical model shows more promising results, although with problems in modelling rapid iceaccretion and ice shedding events.No clear connection between measured production losses and modelled ice loads wasfound when analyzing available data. Uncertainties in input data correction as well asimportance of ice density are possible sources of error.Due to confidentiality of some of the data, the measurement sites used in this thesis aredenoted site A, site B and site C.

As wind power expands rapidly worldwide, it is becoming more common to build wind farms in alpine locations where the wind resources often are good and conflicting interests are few. This is evident in Sweden where a substantial portion of the large wind parks planned are to be built in cold climate locations. The fact that icing of turbine blades and sensors can severely impact the production raises the question how large the losses are. In this thesis 10 wind parks comprising 45 turbines, well dispersed throughout Sweden are investigated. Daily production figures are compared to wind data from the MERRA reanalysis data-set in order to see if it is possible to determine the level of losses during the winter period caused by cold climate. A method is suggested where a relationship between daily production and daily average wind speed is established using representative summer days. This relationship is then used to calculate an expected production for the winter period. Losses are concluded as the difference between expected and actual production. The method did not produce a consistent and reliable result for the sites investigated. However, the method captures the overall trend with higher losses in the north of Sweden compared to the sites in the south where little or no icing is likely. At the sites where icing is expected, losses in the range of 10 to 20% of the annual production were calculated.