Multi-scale Unified Surrogates for Turbulence and Wake Interaction in Next-generation Digital Twins

Descrizione del progetto

Gemello digitale multiscala per l’ottimizzazione dei parchi eolici

I simulatori di parchi eolici trascurano le cruciali interazioni fluido-struttura e la turbolenza di scia, portando a previsioni imprecise della potenza erogata e dei carichi di fatica. Con il sostegno del programma Marie di azioni Skłodowska-Curie, il progetto MUSTWIND svilupperà un gemello digitale multiscala per i parchi eolici di nuova generazione, a sostegno degli obiettivi climatici europei per il 2050. Il progetto utilizzerà simulazioni avanzate e l’apprendimento automatico per generare set di dati dettagliati sulla turbolenza di scia e sulle interazioni fluido-struttura. L’iniziativa mira a consentire il controllo predittivo della produzione e della manutenzione di energia, riducendo al contempo i costi operativi. Combinando il calcolo ad alte prestazioni con la convalida sperimentale, il progetto rappresenta un passo significativo verso l’ottimizzazione dell’energia eolica e l’accelerazione della transizione verso le energie rinnovabili in Europa.

Obiettivo

This project develops a groundbreaking multi-scale digital twin for next-generation wind farms, directly supporting Europe's 2050 climate neutrality goals under the Green Deal. Current wind farm simulators fail to capture critical fluid-structure interactions and wake turbulence dynamics, leading to poor predictions of power output and fatigue loads. This research addresses these limitations by integrating high-fidelity Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS) with two-way fluid-structure interaction (FSI) coupling and machine learning.
The 24-month fellowship at Imperial College London's Turbulence Simulation Group, under Prof. Sylvain Laizet's supervision, pursues four key objectives: (1) generating high-resolution datasets of wake turbulence and FSI under misaligned, intermittent inflows; (2) characterising multi-scale physics through advanced diagnostics including vortex identification and fatigue analysis; (3) developing Bayesian-trained surrogate models for dynamic actuator-line corrections and reduced-order FSI coupling; and (4) integrating these surrogates into a farm-scale solver with real-time optimisation capabilities.
The methodology combines computational excellence with experimental validation, utilising European Tier-0 HPC facilities and wind tunnel data. Three interconnected work packages progress from high-fidelity data acquisition through physical analysis to surrogate model development, enabling predictive control of power output, fatigue loads, and maintenance scheduling.
Expected impacts span scientific advancement (first validated multi-scale wind farm digital twin), economic benefits (reduced O&M costs, improved LCOE), and societal gains (accelerated renewable energy transition, job creation). The project exemplifies interdisciplinary innovation, merging CFD, machine learning, and experimental validation to transform wind energy optimisation, positioning Europe at the forefront of digital twin technologies.