Forecasting Wind Turbine Production Losses due to Icing

Abstract

In order for wind energy producers to avoid the extra costs of inaccurate production estimates for the day-ahead energy market, precise forecasts of power production during the next day have to be made. For renewables such as wind energy, power production is particularly difficult to predict as it depends on the fluctuating wind speed. In addition, forecasting the power production in cold climates such as Norway is further complicated by the icing on wind turbine blades. The extra load on turbine blades due to icing leads to a decrease in power production. Kjeller Vindteknikk has a state of the art model that estimates the production loss due to icing for new wind farms. This model uses historical weather data from the numerical weather prediction model WRF. To meet the interest from customers, Kjeller Vindteknikk wishes to further develop the icing model into IceLossForecast - An operational model providing production forecasts with icing loss for wind energy producers in cold climates. For this to work, the first step is to implement forecasting data into the IceLoss2.0 model and see if icing loss forecasts is comparable to historical IceLoss2.0 data. This is the objective of this thesis. Forecasting data from WRF forecast and the MetCoOp Ensemble Prediction model (MEPS) is implemented, and their results are compared to each other and to the current IceLoss2.0 estimates. In this thesis MEPS and WRF forecasts has been implemented to IceLoss2.0 successfully. Results from one turbine at Kvitfjell Wind Farm has shown promising results both for WRF and MEPS that are comparable to the traditional IceLoss2.0 estimates with historical WRF data.