This document summarizes research on forecasting production losses at a wind farm in central Sweden due to icing conditions. The researchers used observational turbine data, weather research and forecasting (WRF) model output, and an icing model to develop and validate statistical models for predicting hourly wind farm power output. They found that a combination of WRF parameters and icing model outputs best predicted production losses across the three wind parks. Cross-validation showed these statistical approaches improved upon a simple threshold method. Future work will apply the method to additional sites and evaluate forecasts using predicted winds.