A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays

Kiran Perkins; Mahdi Erfanian Nakhchi

doi:10.1109/EFEA67685.2025.11386411

A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays

Kiran Perkins^*, Mahdi Erfanian Nakhchi

^*Corresponding author for this work

School of Engineering Physics and Mathematics

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Downloads (Pure)

Abstract

This study presents a scalable hybrid framework combining Large Eddy Simulation (LES) and Artificial Neural Networks (ANNs) for fast and accurate prediction of yawed wind turbine wakes. High-fidelity LES using the Actuator Disk Method (ADM) was conducted for tandem NREL-5MW turbines under yaw angles of 0°-40° and spacings of 3D-10D. The resulting velocity fields trained spacing-specific ANNs with adaptive yaw-weighted loss functions. The networks reproduced LES wake structures with mean relative errors below 7% while reducing computation time from 17.2 hours per LES run to 0.16 seconds. The model accurately captured wake deflection, recovery, and velocity deficits across configurations, achieving a 400,000× speed-up over LES. This hybrid LES-ANN framework offers a robust and computationally efficient tool for large-scale wind farm optimisation and wake control analysis.

Original language	English
Title of host publication	2025 9th International Conference on Environment Friendly Energies and Applications (EFEA)
Place of Publication	Piscataway, United States
Publisher	IEEE
Pages	1-6
Number of pages	6
ISBN (Electronic)	9798331550950
ISBN (Print)	9798331550967
DOIs	https://doi.org/10.1109/EFEA67685.2025.11386411
Publication status	Published - 4 Dec 2025
Event	9th International Conference on Environment Friendly Energies and Applications, EFEA 2025 - Newcastle upon Tyne, United Kingdom Duration: 4 Dec 2025 → 5 Dec 2025

Conference

Conference	9th International Conference on Environment Friendly Energies and Applications, EFEA 2025
Country/Territory	United Kingdom
City	Newcastle upon Tyne
Period	4/12/25 → 5/12/25

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 7 Affordable and Clean Energy

Keywords

Actuator Disk Method
Artificial Neural Networks
Large Eddy Simulation
Machine Learning
Wake Steering
Wind Turbine Wake
Yaw Control

Access to Document

10.1109/EFEA67685.2025.11386411

KPerkins_submissionAccepted author manuscript, 650 KBLicence: CC BY

Cite this

@inproceedings{954f2cc5bad64ec9b02c0b0dd0f60bba,

title = "A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays",

abstract = "This study presents a scalable hybrid framework combining Large Eddy Simulation (LES) and Artificial Neural Networks (ANNs) for fast and accurate prediction of yawed wind turbine wakes. High-fidelity LES using the Actuator Disk Method (ADM) was conducted for tandem NREL-5MW turbines under yaw angles of 0°-40° and spacings of 3D-10D. The resulting velocity fields trained spacing-specific ANNs with adaptive yaw-weighted loss functions. The networks reproduced LES wake structures with mean relative errors below 7\% while reducing computation time from 17.2 hours per LES run to 0.16 seconds. The model accurately captured wake deflection, recovery, and velocity deficits across configurations, achieving a 400,000× speed-up over LES. This hybrid LES-ANN framework offers a robust and computationally efficient tool for large-scale wind farm optimisation and wake control analysis.",

keywords = "Actuator Disk Method, Artificial Neural Networks, Large Eddy Simulation, Machine Learning, Wake Steering, Wind Turbine Wake, Yaw Control",

author = "Kiran Perkins and \{Erfanian Nakhchi\}, Mahdi",

year = "2025",

month = dec,

day = "4",

doi = "10.1109/EFEA67685.2025.11386411",

language = "English",

isbn = "9798331550967",

pages = "1--6",

booktitle = "2025 9th International Conference on Environment Friendly Energies and Applications (EFEA)",

publisher = "IEEE",

address = "United States",

note = "9th International Conference on Environment Friendly Energies and Applications, EFEA 2025 ; Conference date: 04-12-2025 Through 05-12-2025",

}

Perkins, K & Erfanian Nakhchi, M 2025, A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays. in 2025 9th International Conference on Environment Friendly Energies and Applications (EFEA). IEEE, Piscataway, United States, pp. 1-6, 9th International Conference on Environment Friendly Energies and Applications, EFEA 2025, Newcastle upon Tyne, United Kingdom, 4/12/25. https://doi.org/10.1109/EFEA67685.2025.11386411

A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays. / Perkins, Kiran ; Erfanian Nakhchi, Mahdi.
2025 9th International Conference on Environment Friendly Energies and Applications (EFEA). Piscataway, United States: IEEE, 2025. p. 1-6.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays

AU - Perkins, Kiran

AU - Erfanian Nakhchi, Mahdi

PY - 2025/12/4

Y1 - 2025/12/4

N2 - This study presents a scalable hybrid framework combining Large Eddy Simulation (LES) and Artificial Neural Networks (ANNs) for fast and accurate prediction of yawed wind turbine wakes. High-fidelity LES using the Actuator Disk Method (ADM) was conducted for tandem NREL-5MW turbines under yaw angles of 0°-40° and spacings of 3D-10D. The resulting velocity fields trained spacing-specific ANNs with adaptive yaw-weighted loss functions. The networks reproduced LES wake structures with mean relative errors below 7% while reducing computation time from 17.2 hours per LES run to 0.16 seconds. The model accurately captured wake deflection, recovery, and velocity deficits across configurations, achieving a 400,000× speed-up over LES. This hybrid LES-ANN framework offers a robust and computationally efficient tool for large-scale wind farm optimisation and wake control analysis.

AB - This study presents a scalable hybrid framework combining Large Eddy Simulation (LES) and Artificial Neural Networks (ANNs) for fast and accurate prediction of yawed wind turbine wakes. High-fidelity LES using the Actuator Disk Method (ADM) was conducted for tandem NREL-5MW turbines under yaw angles of 0°-40° and spacings of 3D-10D. The resulting velocity fields trained spacing-specific ANNs with adaptive yaw-weighted loss functions. The networks reproduced LES wake structures with mean relative errors below 7% while reducing computation time from 17.2 hours per LES run to 0.16 seconds. The model accurately captured wake deflection, recovery, and velocity deficits across configurations, achieving a 400,000× speed-up over LES. This hybrid LES-ANN framework offers a robust and computationally efficient tool for large-scale wind farm optimisation and wake control analysis.

KW - Actuator Disk Method

KW - Artificial Neural Networks

KW - Large Eddy Simulation

KW - Machine Learning

KW - Wake Steering

KW - Wind Turbine Wake

KW - Yaw Control

UR - https://www.scopus.com/pages/publications/105035006980

U2 - 10.1109/EFEA67685.2025.11386411

DO - 10.1109/EFEA67685.2025.11386411

M3 - Conference contribution

AN - SCOPUS:105035006980

SN - 9798331550967

SP - 1

EP - 6

BT - 2025 9th International Conference on Environment Friendly Energies and Applications (EFEA)

PB - IEEE

CY - Piscataway, United States

T2 - 9th International Conference on Environment Friendly Energies and Applications, EFEA 2025

Y2 - 4 December 2025 through 5 December 2025

ER -

A Scalable Hybrid LES-ANN Framework for Wake Prediction in Yawed Wind Turbine Arrays

Abstract

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this