Meteorological data sets and wind farm notes were reviewed to identify times when icing conditions were present at wind farm locations in southwestern MN and northeastern CO. The icing events were further analyzed to determine the drop sizes, intensity, and event duration. This resulted in the creation of three categories of icing; cloud icing, freezing drizzle icing and freezing rain icing.

Two case studies were completed for each of the icing categories. The large and small scale meteorological conditions that produced the icing were assessed. The cloud icing events studied appeared to minimally impact the predicted power output. The freezing drizzle events rapidly decreased the power output below the predicted values. The freezing rain events also significantly reduced the observed power output to values well below those expected.

A forecast icing algorithm based on some of the ideas from the Forecast Icing Product (FIP; McDonough et al., 2004) was developed to identify each of the three categories of icing. Detection of the cloud base height, above or below turbine hub height, along with wind speed and event duration time are important to the cloud icing category. Differentiating between snow clouds and freezing drizzle clouds, as well as between only cloud icing compared to freezing drizzle icing are important factors in the freezing drizzle icing category. The prediction of a deep precipitating cloud, along with a subsequent melting layer aloft, and a surface based sub-freezing layer were key to identification of the freezing rain icing category. The algorithm was tested at the two farm locations for each of the cases. This paper shows the icing conditions observed, presents the case studies, shows the resulting power output, and presents the icing detection algorithm for each of the cases.