Short-term prediction of icing-related production losses
The icing of turbine rotor blades is one of the most common causes for large-scale production losses in wind farms in regions with mountainous terrain. Depending on site and altitude, these losses can amount to 10% worth of yearly production. Also, unexpected icing-related wind turbine shutdowns can entail significant balancing costs due to the mechanisms of the energy market and power grid regulations.
To predict atmospheric icing conditions and their impact on a turbine’s rotor blades represents one of the biggest challenges of energy meteorology.
Austro Control Digital Services has partnered with energy traders to develop icing predictions that combine state-of-the-art numerical weather prediction with machine learning algorithms. High-resolution ensemble weather models are fused with current turbine data and cloud measurements and are calibrated based on historical icing episodes.
The resultant predictions with a short-term horizon (3-4 h) and frequent updates (15 min) are tailored to be used on intra-day energy markets. They empower traders to act quickly before and during turbine icing, thereby avoiding significant balancing costs.
Hard facts
- Icing on rotor blades can lead to significant production losses
- Prediction of icing is a huge challenge
- Numerical weather prediction data are fused with current SCADA data and cloud measurements and calibrated using historical icing episodes
Benefits
- Short-term heads-up warnings for intra-day traders
- Perfecting all relevant input data streams through state-of-the-art methods of machine learning
- Enabling quick counter-measures in adverse icing situations
- Avoiding significant production losses and balancing costs
