Presentations

Abstract

In the context of increasing onshore wind capacity worldwide, with much of this development occurring in cold climates, understanding the impact of icing on energy production has become increasingly important. While cold climate regions such as the Nordics have long been a focal point for addressing ice accretion on wind turbines, there is a growing need to extend this understanding to other regions with diverse cold climatic conditions and to improve and validate industry tools that can cover all market regions. DNV has utilized the WICE 2.0 model to estimate onshore wind production losses. This model chain integrates site-specific mesoscale modeling using the WRF model, ice accretion modeling for specific turbine types, and a machine learning model trained on high-resolution operational data from wind farms located in harsh cold climate conditions in the Nordics. This approach allows the translation of project-specific climate conditions modeled with WRF, project-specific ice accretion, and, when available, local measurements into ice production loss estimates and long-term correction methods for model results. The current WICE 2.0 model has been validated in the Nordics and applied across diverse climatic and operational contexts worldwide, including the USA, Canada, Central Europe, the Balkans, and some countries in the Middle East, demonstrating strong performance. Building on these experiences, DNV has introduced the new WICE 2.5 model, an update from WICE 2.0, and extended its validation using atmospheric, terrain, and operational data from non-Scandinavian regions to improve accuracy in diverse locations. This includes data from various representative sites in North America and Northern Europe, extending the prediction capabilities of WICE 2.5 to distinct climatic, operational, and topographic conditions. Additionally, the forcing WRF dataset was updated from ERA-Interim to ERA5. Results showed that training the model with region-specific operational, climatic, and topographic data led to more accurate predictions of energy losses due to icing compared to the previous model trained solely with Nordic data. Comparisons and validations were conducted using IEA Task 19 methodologies and DNV's methods to estimate icing losses. Consequently, the uncertainty in loss predictions can be reduced when the model is trained with data representative of the intended application region. The model can also continuously improve based on further validation. Another strength of the new WICE 2.5 model is its capability to incorporate the effects of de-icing, anti-icing systems, or icing control modes, provided adequate system information is available. Furthermore, the WICE 2.5 model can operate in a different mode to account for icing shutdowns required for administrative or safety reasons. This presentation will introduce the new modeling approach, present validation results, and offer an outlook for ongoing and future validation efforts to ensure the model's applicability across global markets.