A surrogate-assisted reinforcement learning framework for layout optimization and economic assessment of dense urban vertical-axis wind turbine arrays

Urban wind energy is limited by space constraints and wake losses in dense arrays. This study integrates CFD-informed surrogate modelling with reinforcement learning to optimize compact vertical-axis wind turbine (VAWT) layouts. A three-stage clustering method reduces 82,103 complete hourly records from Winnipeg (2016–2025) to four representative speed states. A 14-component Gaussian mixture identifies joint speed–direction regimes; k -means extracts 12 centroids, which are aggregated into four weighted speeds (2.99, 3.18, 6.25 and 10.12 m s -1 ) covering the full record. A five-layer neural network surrogate, trained on RANS simulations across speeds and spacings, predicts power with less than 5 % error relative to CFD and manufacturer data. Embedded in a Deep Q-Network, the surrogate optimizes placement in a 10 m × 10 m domain, yielding a 45-turbine layout that produces 4.8 kW at 26.6 % efficiency and 48.0 W m -2 . Techno-economic assessment indicates near-term feasibility for decentralized urban wind energy.

Authors

Institutions

Publication Details

Journal
Wind Engineering
Published
2026-11-16
DOI
https://doi.org/10.1177/0309524x261506803
Primary Topic
Wind Energy Research and Development
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A surrogate-assisted reinforcement learning framework for layout optimization and economic assessment of dense urban vertical-axis wind turbine arrays

Ali A. Zaidi
2 links
Wind Engineering
Wind Energy Research and Development
article

A surrogate-assisted reinforcement learning framework for layout optimization and economic assessment of dense urban vertical-axis wind turbine arrays

Ali A. Zaidi
article en

Abstract

Urban wind energy is limited by space constraints and wake losses in dense arrays. This study integrates CFD-informed surrogate modelling with reinforcement learning to optimize compact vertical-axis wind turbine (VAWT) layouts. A three-stage clustering method reduces 82,103 complete hourly records from Winnipeg (2016–2025) to four representative speed states. A 14-component Gaussian mixture identifies joint speed–direction regimes; k -means extracts 12 centroids, which are aggregated into four weighted speeds (2.99, 3.18, 6.25 and 10.12 m s -1 ) covering the full record. A five-layer neural network surrogate, trained on RANS simulations across speeds and spacings, predicts power with less than 5 % error relative to CFD and manufacturer data. Embedded in a Deep Q-Network, the surrogate optimizes placement in a 10 m × 10 m domain, yielding a 45-turbine layout that produces 4.8 kW at 26.6 % efficiency and 48.0 W m -2 . Techno-economic assessment indicates near-term feasibility for decentralized urban wind energy.

Wind Engineering
University of Manitoba (CA)
Affordable and clean energy, Industry, innovation and infrastructure, Climate action, Sustainable cities and communities
Wind Energy Research and Development
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon