VALEMO/Météo France partnership for the development of an ice forecasting model

Modelling of ice accretion and melting on wind turbine blades

VALEMO operates several parks located in areas regularly impacted by icing phenomena. The consequences of icing on wind turbine blades concern, on the one hand, significant production losses and, on the other hand, induce premature ageing of the wind turbines in case of operation with a load of ice on the blades and also imply a problem on the safety of people in the vicinity of the wind farm.

As can be seen in Figure 1, a blade can be covered with a thick layer of frost. In these conditions, if it starts to rotate, blocks of ice may be thrown around at tens or even hundreds of meters. The challenge is therefore to minimize downtime due to icing while ensuring safety in the park. VALEMO is participating with Météo France in the development of a forecast model for ice accretion and melting on wind turbine blades.

Météo France has a tool called WIRE (Winter Risk for Energy) dedicated to alerting players in the energy sector to winter weather risks. This tool was mainly applicable to the modeling of frost accretion on overhead power transmission cables. The modeling is made from the fields of the high-resolution weather forecasting model AROME. Thanks to the participation of Valemo, the WIRE model of Météo France has been modified in order to adapt it to the wind sector (different obstacle dimensions, taking into account the rotation of the blades, etc). Then, a melting model was developed in order to have a forecast followed by the icing event, from accretion to disappearance.

A validation of the ice accretion model was carried out in 2019 using data from the VALEMO remote control center and operational data from a power plant in the Montagne Noire region, which is heavily impacted by ice. The very convincing results obtained led us to renew the partnership between Météo France and VALEMO in 2020 to validate an ice melting model. As operator, the objective is both to detect the onset of icing episodes in order to limit the mechanical load on the turbines, and to detect the end of the episode in order to optimize the production of the fleet without generating a risk for people and installations. Validating this type of model is always very complex, since operators like VALEMO have little way of knowing precisely when icing and melting occurs in their fleets, hence the purpose of the work with Météo France. Two wind turbines of this wind farm operated by VALEMO, located in the Montagne Noire, were instrumented in November 2019, however the winter 2019-2020 was particularly mild and no icing episode has been recorded since the instrumentation. The validation was therefore carried out using two methods:
– Use of the Météo France test site located at the Col de Porte (Isère, Massif de la Chartreuse). This site has cameras aimed at transport cables. It is therefore possible to know precisely the dates-time of accretion and melting on these cables.
– Use of historical data from shutdown periods and restart of the wind turbines of the VALEMO fleet. However, the accuracy of the dates is low.

Figure 2 shows simulation results obtained with different models (different assumptions in particular). The dotted black vertical line shows the beginning of frost accretion and the solid black vertical line indicates the end of the melting (total disappearance of frost). The different results showed that simulations 4 and 6 allowed to represent well the dynamics of the different events observed at the Porte pass. For example in figure 2, we observe the phenomena of stabilization of the melting during the night. Simulation 6 has the advantage of a hypothesis on the infrared radiation that allows the model to get rid of a parameter that is difficult to calculate, namely the infrared radiation due to the ground (depending in particular on the topology of the terrain and the nature of the soil).

Figure 3 shows the results obtained on a fleet of the VALEMO fleet during the winter 2018/2019. The AROME ensemble forecast was used (the aim is to predict the appearance of icing 12 hours in advance). We can see that a very good detection score is obtained for probability thresholds between 10 and 40% with 5 episodes detected out of 7. We thus obtain a detection rate of 78.6%. The associated false alarm rate is rather low, calculated at 13.4%.

The first results obtained are promising and VALEMO and Météo France will continue their partnership in order to continue the validation of the model on other observational data. If next winter is less clement, we will be able to analyze precisely the data from icing sensors installed on 2 wind turbines. Work is also planned on the adaptation of the operation of the wind farms according to the forecasts provided by the WIRE model.

References