Pedersen, Marie Cecilie

Industrial PhD Vattenfall Vindkraft A/S

Project title: WINDICE: Reliable Prediction of Icing on Wind Turbines and Icing-Induced Power Loss

PhD period: 2014.01.01 - 2016.12.31 (has been prolonged till 2018.04.01)
Section: Fluid Mechanics and Combustion
Research Programme: Wind Power Systems
Supervisors: Henrik Sørensen, ET, AAU and Annika Billstein, R&D Engineer at Vattenfall in Sweden
Co-Supervisor: Thomas Condra
Contact Information

Collaborator: Vattenfall Vindkraft A/S.
Funding: Vattenfall Vindkraft A/S, Ministeriet for Forskning, Innovation og Videregående Uddannelser. 


The continued expansion of wind power in regions exposed to icing is heavily dependent on improved understanding of the relation between icing on wind turbines and the resulting production losses and safety risks. For Vattenfall’s wind farm at Stor-Rotliden in Sweden the annual production losses due to icing are on the order of 3-15 % of the total production. Losses of this magnitude have a huge impact on the profitability of a wind farm. So without a better ability to predict icing-induced losses and/or ways to reduce these, the commercial consequences may be devastating. Right now, Vattenfall is developing four wind farms in cold climate, which is expected to add about 250MW icing-affected capacity by 2015. With an assumed average ice-induced production loss of approximately 8% of the annual energy production the research on ice accretion is crucial.
Thus, cold-climate wind energy is challenged by icing on the wind turbines and a series of icing-induced problems such as production loss, blade fatigue, and safety issues due to ice shedding. Because of the difficulties with on-site measurements, simulations are often used to understand and predict icing events. The current efforts on ice accretion forecast are mainly based on two different modelling concepts.

One is the ice accretion model proposed by Makkonen (1984), which describes in-cloud icing on a vertically placed, freely rotating cylinder. The ice accretion rate calculated by the model mainly depends on three efficiencies: droplet collision efficiency, sticking efficiency and accretion efficiency. This model is commonly used for forecasting icing events in wind farms. Figure 1 illustrates how Makkonen’s model is incorporated in the power loss assessment methodology currently employed by Vattenfall’s Turbing Icing Program. However, one must be aware of the shortcomings of the model when it is applied in wind power. Icing characteristics on a vertical, freely rotating cylinder are most likely different from icing on a wind turbine blade in circular motion in a vertical plane as the shape, dimensions and motion style are very different. So far, there is no algorithm to convert the ice load modeled on a cylinder into an ice load on a wind turbine blade.


Figur 1. Current methodology for prediction of production loss due to icing.

The other conceptual model is dedicated for simulation of ice accretion on two-dimensional airfoils, which are mostly seen by panel based methods or recently by Computational Fluid Dynamics. But until now, none of the panel-based or CFD-based methods has been coupled with numerical weather models to address variations in external meteorological conditions during an icing event.

The advantage of ice accretion models implemented in CFD is that the ice-induced changes in the aerodynamic properties can be determined. This is paramount when establishing the most-requested relation between icing events and the resulting loss of power production, as seen in the Empirical Power Model in Figure 1. For Vattenfall it is thus of great interest to develop a CFD model to reliably simulate the whole icing event on a wind turbine blade, based on which an engineering model for icing-induced power loss can also be derived.

Research hypothesis

The PhD project includes the following hypotheses.
• A general CFD model applicable to ice accretion on any structure can be developed. It can utilize a dynamic mesh for grid-displacement during icing and real-time series of external meteorological conditions. This can be obtained by integrating the CFD inputs with appropriate models like a k-ω model for clear air flow, an Eulerian multiphase model for droplet flow, Messinger’s model for ice accretion, and dedicated models to be proposed in this project for ice melting, shedding and sublimation effects.
• The applicability and reliability of the general CFD model can be experimentally verified by data collected from the ice sensor at Stor-Rotliden wind farm. Further validation is possible by using experimental data from lab-scale wind turbine blades.
• From sensitivity analysis based on the validated general CFD model, an in-depth understanding of icing events can be achieved, key parameters in the icing process can be identified, and their impacts can be better quantified.
• A reliable engineering model to predict icing-induced production loss from e.g. wind velocity and ice load can be derived from the general CFD model.

The project is to better understand the fundamentals of ice accretion on a wind turbine and to more reliably predict operational production losses in areas of icing risks, which can be used to improve spatial planning of new wind farms in cold climate regions. Therefore, developing a general CFD model for ice accretion, getting the general model validated, extracting knowledge from the validated model via sensitivity analysis, and deriving an engineering model for practical use are the core hypotheses/questions which are very closely related to the project objectives.

Project methodology and anticipated results

The overall methodology for predicting icing events and icing-induced production loss to be developed in this project is sketched in Figure 2. The anticipated results of the PhD can be divided into two parts. One is to develop a general CFD-based ice accretion model and the other is to propose a new engineering model for icing-induced power loss prediction. The other one is to develop a new, promising alternative for icing-induced power loss prediction, which is shown in Figure 2.

Figur 2. Methodology for prediction of icing-induced production loss, to be developed in this project.

In final goal of the project is to introduce a reliable model to forecast production losses due to icing, which will reduce the financial risk in these investments. Knowledge about icing and production loss characteristics of the individual wind turbine will be used to optimize the layouts of future wind farms, and to optimize production and maintenance of existing wind farms. The project will furthermore contribute to a more effective operation of existing wind farms in cold climate regions and enable estimations of the required energy consumption of heating solutions for blade de-icing.


Publications in journals and conference papers may be found at VBN.