Posters

Abstract

In icing climates, ice can form on wind turbines when supercooled liquid droplets in clouds freeze on turbine blades. This in-cloud icing leads to production losses due to changes in aerodynamics as well as the need to stop turbines as a safety procedure due to ice throw. In the Nordics, icing losses are the second highest losses after wake losses. Despite the global warming, they can be expected to remain so in the near future with the technology trend of building bigger wind turbines closer to low level clouds. Forecasting icing correctly would benefit balancing of the electricity system as well as single wind energy producers. However, to date, turbine icing has been extremely difficult to forecast due to its dependence on modelling the horizontal and vertical location of humidity and clouds correctly. We at Kjeller Vindteknikk have developed a physical icing forecast model, IceLossForecast, for day-ahead applications based on +10 years of experience of icing modelling for project development. IceLossForecast calculates the icing on turbine blades and subsequent production losses for each hour for the next 48 hours. The model uses our in-house mesoscale WRF (Weather Research and Forecasting) forecasts as input data. To assess uncertainties and the probability of icing, IceLossForecast includes ensemble simulations by altering, for instance, the cloud base height and the location and timing of clouds. Furthermore, the model can utilize real-time SCADA data to continuously recalibrate the modelled ice mass on turbine blades. KVT has produced operational icing forecasts for all wind farms in Finland for a pilot period in winter 2023-2024 for the Finnish transmission system operator Fingrid to improve their grid balance forecasts. Furthermore, project level icing forecasts for several wind farms have been produced with and without real-time SCADA information. In my presentation, I will discuss the performance of the forecasts and the challenges encountered.