AFRY DAPS

Predicting icing loss and energy production

The complex nature of the icing phenomenon and its connection to various weather conditions has led many to believe that it is impossible to predict the production effects of all icing events correctly. We think otherwise. As a result of AFRY’s development work, production losses caused by turbine icing can now be predicted significantly better than before.

AFRY’s prediction model enables a data-based approach to solve the problems of predicting icing events: by studying the relationship between wind farm production data and weather conditions, the prediction model is optimized for each wind farm separately.

This allows the consideration of different blade surface materials and shapes affecting how ice is accumulated and released from the turbines.

Prediction errors during icing events are among the most significant sources of additional costs in wind farm production. According to AFRY’s study, the prediction model can reduce forecast errors caused by icing significantly and almost all the icing events can be predicted using the model. This can reduce the annual costs of prediction errors by thousands of euros per turbine.

Challenges related to icing

• In cold climate regions, icing on wind turbine blades significantly impacts the operation of wind farms.
• Icing losses can cause large uncertainties in production forecasts.
• Regulating prices in energy markets are typically very high during icing conditions, emphasizing the financial impact of prediction accuracy.

AFRY’s icing model compared to other models on the market

Compared to many other icing prediction models on the market, AFRY’s icing model:

• Considers different icing types in the modelling (clear ice can cause even worse production losses compared to typical rime ice).
• Focuses on point of ice release, which usually has the most significant impact on event costs.
• Model parameters are optimized based on the SCADA history data of each wind farm.
• Model is applicable for turbines with ice prevention systems or other specifications.

Using AFRY DAPS in a wind farm

The system integrates weather forecast information, in-house models, SCADA data, and advanced data analysis techniques. Power output predictions are automatically calculated from 15 minutes to 10 days in advance, for individual turbines, wind farms, or entire portfolios. The system includes turbine-specific production and icing loss estimates and helps to avoid higher balancing costs and imbalanced prices. SCADA data, including wind farm production, wind speed, wind direction, and temperature, is combined with weather forecast data to achieve the most accurate predictions.

Model accuracy

The model is based on ice accretion physics and data analysis techniques. Model validations have been performed using historical data from various wind farms and historical energy market pricing. The validation results indicate that the AFRY DAPS icing model can significantly reduce prediction error costs. From current client projects utilising AFRY DAPS, the annual potential cost reductions of using the icing model would have been tens of thousands of euros per wind farm. According to the results from these projects, the model has found and predicted about 80% of the icing events during wintertime.