Norsk English

PreWinT - Prediction and Mitigation of Wind-Turbine Noise and Its Impact on Humans

WP5 - Wind farm noise model (digital twin)

This work package aims to refine wind farm operational strategies by developing an advanced digital twin for wind turbine noise management. The goal is to support adaptive noise reduction strategies that comply with regulatory noise limits while minimizing power output curtailment.

Contact person

Building on advancements in modelling of flow field (T2.3), wind turbine noise generation (WP3), and noise propagation (WP4), the project will construct a comprehensive wind park noise model to simulate and predict noise impacts. To address different operational needs, two scenarios will be evaluated during the initial design phase:

  • The Forecasting Scenario approach uses weather forecasts and planned operational data to simulate wind fields, enabling the physical model to predict noise generation and propagation over longer time horizons.
  • Alternatively, the Nowcasting Scenario approach relies on real-time measurements of wind fields and weather conditions (from local metrology and LIDAR), supplemented by acoustic data, to dynamically calibrate noise predictions for immediate operational adjustments.

Because comprehensive direct measurements across extensive spatial domains are impractical, data-driven machine learning (ML) models will be trained on simulation outputs from these physics-based models. A promising hybrid approach involves using a simplified physics-based model as a baseline and applying an ML-trained correction function to account for residual errors and unmodelled phenomena46.

Success is defined by the ML models’ ability to reliably reproduce the physical 5 model’s outputs with enhanced computational efficiency. A rigorous evaluation—based on predictive accuracy, computational efficiency, and stakeholder feedback—will be conducted during the design phase (T5.1) to select the most feasible and valuable scenario (forecasting or nowcasting). The chosen scenario will then inform the development of the subsequent digital twin components (physics-based, ML-based, and hybrid)

This will be done through the following tasks: 

  • T5.1 Design Specification and Feasibility Evaluation 
  • T5.2 Development of the Physics-Based Model/Digital Twin 
  • T5.3 Development of a Data-Driven ML-Model/Digital Twin 
  • T5.4 Development of a Hybrid Model/Digital Twin 
  • T5.5 Evaluation of Model/Digital Twin Performance