ABSTRACT This paper proposes a novel genetic algorithm‐based proportional–integral super‐twisting sliding mode control (GA‐PI‐ST‐SMC) strategy for the rotor‐side converter of a 1.5 MW doubly‐fed induction generator (DFIG) operating in marine renewable wind turbine (MRWT) systems. The proposed approach integrates the strong robustness and finite‐time convergence characteristics of super‐twisting sliding mode control with the global optimization capability of genetic algorithms to achieve optimal controller tuning and enhanced system performance under varying operating conditions. The effectiveness of the proposed controller is systematically evaluated under both nominal and perturbed conditions and benchmarked against the conventional super‐twisting sliding mode controller (ST‐SMC). Comparative results reveal substantial improvements in power quality and dynamic response. Under nominal operating conditions, the proposed GA‐PI‐ST‐SMC reduces active power ripple by 46.35% and current total harmonic distortion (THD) by 34.60%. Moreover, under robustness tests involving parameter variations and system uncertainties, the controller maintains superior performance, achieving reductions of 95.06% in reactive power ripple and 39.14% in current THD compared with the conventional ST‐SMC scheme. These results demonstrate that the proposed optimization‐based control framework significantly enhances disturbance rejection capability, suppresses power oscillations, and improves current waveform quality while preserving system stability and robustness. Consequently, the GA‐PI‐ST‐SMC emerges as an effective and reliable control solution for high‐power DFIG‐based MRWT applications, contributing to the advancement of robust and high‐efficiency renewable energy conversion systems.
ABSTRACT This study investigates the consistency and impact of feature selection methods on deterministic and probabilistic normal behaviour models (NBMs) for estimating rear generator bearing temperature in wind farms using SCADA data. Rigorous NBM development enables the transfer of key features, enhances anomaly detection comparability and facilitates potential portability across wind farms. The rise of machine learning and a vast range of statistical methods available for anomaly detection or estimation with SCADA data further emphasizes the need for investigating consistent NBMs. This paper evaluates four feature selection methods—Pearson's correlation coefficient (PCC), decision tree weights (DTW), mutual information (MI) and Shapley values of a neural network (SHAP)—for two wind farms to assess consistency in identified influential features and their impact on estimation algorithms. The results highlight the importance of prioritizing features consistently identified across datasets and methods over merely optimizing deterministic estimation performance. Temperature‐related sensors dominated the key features, but their specific locations were also ranked consistently. Probabilistic interval estimations demonstrated superior estimation performance and to identify whether a significant difference in feature can be attributed to the model or to an anomaly in the measured data.
ABSTRACT Offshore wind energy development in China's coastal regions represents a critical pathway toward achieving carbon neutrality goals, yet accurate resource assessment remains challenging due to complex marine meteorological conditions. This study presents a comprehensive evaluation of offshore wind resources in Fujian Province using convection‐permitting Weather Research and Forecasting (WRF) model with 4‐km grid spacing nested in a 12‐km grid, validated against 1 year of buoy measurements and ERA5 reanalysis data. Our high‐resolution simulations reveal superior performance in capturing both routine and extreme wind conditions, with particular improvement in the critical 5–20 m/s wind speed range crucial for wind power generation. Analysis of 22 years of simulated data uncovers distinct seasonal and spatial patterns in wind resources: winter months contribute significantly higher wind power density, while the Taiwan Strait emerges as an optimal development zone with strong, stable winds and notably moderate extreme conditions. Extreme value analysis indicates that maximum wind speeds in the Taiwan Strait remain below 31 m/s during 100‐year return periods, substantially lower than surrounding areas, suggesting reduced structural requirements for wind installations. These findings provide crucial guidance for offshore wind farm siting and design, highlighting the Taiwan Strait as a prime location combining excellent wind resources with manageable extreme event risks. Our methodology and results establish a robust framework for offshore wind resource assessment in complex coastal environments, directly supporting China's ambitious renewable energy goals.
ABSTRACT Due to their superior load density, planetary gear journal bearings (PGJB) are progressively replacing rolling element bearings in wind turbine gearboxes; therefore, accurately revealing the dynamic characteristics of drivetrains supported by these bearings under variable operating conditions is crucial for ensuring operational reliability. However, the nonlinear mechanism governing the influence of lubrication state transitions on system vibration response remains unclear. To address this, this paper establishes a flexible–rigid coupled dynamic model of a high‐power wind turbine drivetrain (WTD), aimed at investigating the coupling mechanism between lubrication state evolution and system vibration characteristics. A unified nonlinear force formulation integrating asperity contact and hydrodynamic oil film forces (OFF) is proposed, enabling the continuous simulation of the entire evolution process across boundary, mixed, and hydrodynamic lubrication regimes. The model is validated using field measurement data from an 8 MW wind turbine gearbox. The results indicate that the lubrication state dictates the system's spectral signature: Boundary lubrication induces strong high‐order nonlinear harmonics and modulation sidebands due to solid contact, whereas hydrodynamic lubrication effectively attenuates vibration energy and stabilizes spectral content. Furthermore, the sensitivity of the system dynamic response to variations in rotational speed is found to be significantly higher than to load variations. This study provides theoretical support for the optimization of lubrication strategies and vibration control in high‐power wind turbine gearboxes.
ABSTRACT Accurate wind speed forecasting is indispensable for the efficient integration of wind power into the electrical grid, optimizing turbine scheduling, and ensuring grid stability. Existing models, such as Long Short‐Term Memory (LSTM) networks, often struggle to capture the highly nonlinear relationships and long‐range dependencies present in wind speed data. To address these limitations, this study proposes a hybrid model, LSTM‐FDA‐KAN, which integrates the sequential learning strength of LSTM with the nonlinear approximation capabilities of the Kolmogorov–Arnold Network (KAN). The model is further enhanced by three specialized modules. The Feature Cross (FC) module captures complex interactions between input variables, the Dynamic Residual (DR) module adaptively corrects prediction errors in real time, and the Adaptive Grid (AG) module enables flexible spatioemporal discretization, making the model more robust to uneven data distributions. We evaluate the model using the LandBench 1.0 dataset, which combines diverse meteorological and land‐surface variables across global regions. LSTM‐FDA‐KAN consistently outperforms baseline models, achieving Pearson correlation coefficients nearly 0.91 and reducing root mean square error (RMSE) by 2%–3%. Gains are most pronounced in coastal and high‐latitude regions, where prediction accuracy improves by up to 15% under highly variable weather conditions. Beyond overall accuracy, the model demonstrates stability in capturing long‐range temporal dependencies. It performs reliably during prolonged extreme events, such as consecutive heatwaves and extended rainfall periods, where conventional models typically degrade. Together, these results establish LSTM‐FDA‐KAN as a scalable and effective solution for geophysical time‐series forecasting. By combining sequential modeling, nonlinear function approximation, and adaptive modules, the framework advances both the accuracy and reliability of wind speed prediction.
ABSTRACT As the wind energy industry matures, inspection of wind turbine blades (WTBs) is shifting from a manual process involving rope access and grading of damage, towards unmanned aerial vehicle (UAV) photography and artificial intelligence‐aided processing of data. Damage detection and classification methods have been proposed, using object detection based on convolutional neural networks (CNNs) to localise and quantify damage in images of WTBs. However, existing approaches are unable to measure the detected instances of damage, which is key for effective maintenance planning. A major limitation to crack measurement in images is the requirement for an explicit scale reference marker or additional data such as imaging distance, which is not always practical or cost‐effective in the inspection of large WTBs. Here, we present a novel crack detection and measurement framework for WTBs which does not require additional sensor data, using a fine‐tuned CNN‐based object detection and segmentation network to locate cracks and implicit scale references in images. We detect WTBs within images and then identify cracks, scaling them using known dimensions of the WTBs. This work also provides a comprehensive dataset of representative WTB damage data including images of small‐scale WTBs with simulated cracks, leading edge erosion and contamination. Our approach shows strong performance on the test dataset of images and can detect and measure cracks in limited orthomosaic views produced from 3D reconstructions, indicating that the approach could be applied to the inspection of full‐scale WTBs in the future.
ABSTRACT With the development of ultra‐compact wind turbines, the electromechanical coupling dynamic characteristics of integrated gearbox‐generator systems in their drive trains have attracted increasing attention. To reveal these characteristics, an electromechanical coupling dynamic model is established using the lumped parameter method and the magnetic field energy method, incorporating the internal and external excitations of the gearbox as well as the nonlinear electromagnetic excitation of the generator. To achieve efficient and accurate simulation, this paper proposes a Multi‐scale Fast‐Slow Hybrid Solution Method (MFS‐HSM) based on the time‐scale separation strategy. The proposed method effectively alleviates the conflict between computational accuracy and efficiency in multi‐scale dynamic analysis; compared with the traditional single‐step method, the simulation efficiency is improved by 10–15 times, thereby filling the research gap in multi‐scale solution approaches for ultra‐compact wind turbine drive trains. The dynamic characteristics of the rigidly integrated gearbox‐generator system are analyzed in detail. The results demonstrate a strong unidirectional coupling effect: Fluctuations in generator torque significantly affect gear contact stress. Meanwhile, compared with a standalone gearbox system, the rigid electromechanical coupling structure reduces vibration frequency, alters the dominant vibration mode, and increases vibration amplitude—rendering the system more prone to resonance. These findings provide a solid theoretical basis for the structural design and vibration suppression of ultra‐compact wind turbines.
ABSTRACT Accurate, time‐resolved installed capacity data are crucial for forecasting and analysing wind‐power production. The time series of installed capacity is often only approximately known in regions with rapid wind power development. Public wind power databases may only be updated yearly and installation dates may not be differentiated within wind farms. Errors in the magnitude or timing of added capacity can significantly impact the analysis of a production time series. The cumulative maximum of the power production can be used as a robust approach to estimate installed capacity but requires monotonically increasing capacity and relies on frequent high wind events. Limited information exists in the literature on advanced techniques to determine the installed capacity time series from measured data for a region. Here, we present a method to find the most probable time series of installed capacity for some measured wind power production, using a simulation of capacity factor time series and a quadratic optimization. Results show a 27.2% reduction in normalized mean absolute error when quantifying the installed capacity after a new wind farm is connected. It is also shown that the day‐ahead forecast mean absolute error and root mean square error are reduced by 2.0% and 2.3%, respectively, when using the method to normalize production data before training of a forecasting model, compared to using the cumulative maximum for normalization. An advantage of the proposed method is that it can be used without the assumption of monotonically increasing installed capacity. It is also computationally cheap and thus suitable for usage in automated forecasting pipelines.
ABSTRACT With the rapid expansion of China's wind power industry, the local climatic impacts of large‐scale wind farms, particularly in arid and semi‐arid regions, have attracted increasing attention. This study investigates the effects of the Huitengxile wind farm in Inner Mongolia, which hosts over 1,200 wind turbines, using high‐resolution numerical simulations conducted with the Weather Research and Forecasting (WRF) model. The analysis quantitatively evaluates the effects of the wind farm's influence on surface temperature, surface relative humidity, turbulent kinetic energy (TKE), and wind speed at turbine height by comparing control and sensitivity simulations. The results indicate that the Huitengxile wind farm induces a decrease in surface temperature but an increase in surface relative humidity. At hub height, TKE increases, while wind speed decreases within the wind farm area. These climatic responses exhibit seasonal and diurnal variability. In winter, the surface temperature decreases by approximately 0.5°C, wind speed reduces by 4.4 m/s, and TKE increases by 1.2 m2/s2. In contrast, summer responses are less pronounced, with surface temperature decreasing by 0.2°C, relative humidity increasing by 0.6%, wind speed reducing by 2.0 m/s, and TKE increasing by 0.7 m2/s2. Diurnal variations reveal stronger cooling, wetting, and wind speed reduction during nocturnal hours, whereas TKE enhancement is more significant during diurnal hours. Spatial analysis reveals that the wind farm's impacts are primarily confined to areas with high turbine density. These findings reveal the complex local climatic effects of large‐scale wind farms, providing critical scientific insights for optimizing wind farm planning.
ABSTRACT This study investigates the contact stress and rating life of a three‐row pitch bearing for wind turbines, considering the effects of roller crowning. All crowning profiles were logarithmic, and a total of six profiles were examined based on the Romax, ISO 16281, and Fujiwara equations. Finite element analysis (FEA) was employed to evaluate the contact stress in the bearing elements induced by different crowning profiles. To improve numerical convergence and accurately calculate contact stresses, the rollers in the full bearing model were replaced with spring elements. The reaction forces obtained from the full model were then mapped onto a sub‐model, where detailed contact stress analyses of the rollers were performed. The contact stress magnitudes for each profile were compared under extreme load conditions. For fatigue loading conditions, the modified rating life was calculated in accordance with ISO 16281 using the reaction forces derived from the spring elements. The stress analysis results indicated that the ISO 16281 profile exhibited no edge stress concentration and yielded the lowest stress at the roller contact because it mathematically induces a uniform load distribution. In parallel, the fatigue life calculation showed that the bearing life was longest for the ISO 16281 profile, followed by the Fujiwara and Romax profiles. This indicates that a longer life was achieved with the ISO 16281 profile rather than the Fujiwara profile, despite the latter having a lower roller reaction force distribution per bearing angle. These findings demonstrate that both the overall bearing reaction force distribution, local contact stresses, and bearing life should be comprehensively considered when selecting crowning profiles for wind turbine pitch bearings.
ABSTRACT This study investigates the influence of inflow turbulence generation simulation methods and blade pitch control strategies on the performance and blade structure loading of a large offshore wind turbine. Two inflow representations were considered: synthetic Kaimal turbulence generated in accordance with IEC design standards and large‐eddy simulation (LES) of a neutral atmospheric boundary layer with a mean hub height turbulence intensity of approximately 6%. Both inflows were applied using Taylor's frozen turbulence hypothesis and included steady induction from precursor blade element momentum theory (BEMT) simulations for turbine–flow interaction, enabling a direct comparison of inflow and control effects. The turbine is also represented in the model using a quasi‐steady BEMT code, Maya, which provides aerodynamic forces on to the device. Control strategies evaluated included collective pitch control (CPC) and individual pitch control (IPC) implemented within a proportional‐integral (PI) controller. Results show negligible differences in mean power output between CPC and IPC, indicating that IPC can be employed as a blade fatigue mitigation measure without compromising energy capture. Across both inflow types, IPC reduced the fatigue loads due to the blade root bending moment (RBM) by 11%–15%, with a greater reduction for the Kaimal turbulence. Compared to Kaimal, LES inflow produced lower overall RBM fatigue loads at the same onset wind speeds, despite exhibiting greater average blade‐to‐blade variation due to spatially varying turbulence structures. These spatial asymmetries, which were not fully mitigated by the fixed‐gain IPC, are of particular concern for floating platforms, as they can lead to unbalanced rotor thrust and potential excitation of platform surge and pitch modes. The findings indicate that IEC‐based synthetic turbulence may overpredict fatigue loads relative to more realistic offshore inflow conditions captured by LES, but that spatial variability in such inflows can challenge current control approaches.
ABSTRACT Wind power is a growing source of energy generation that relies on complex global atmospheric and earth system processes. There has been evidence of reductions in average wind speeds over land in North America since the 1980s, and several models project that average wind speeds will continue to decrease. Concurrently, the cost of wind energy systems in the United States has been decreasing since around 2010, a trend also projected to continue. There is considerable uncertainty in these future projections, with quantitative estimates of future wind resource and system costs varying widely. To study this, we run wind energy models with possible future system costs, turbine designs, and meteorological inputs from multiple downscaled earth system models over the contiguous United States. Changes in mean annual energy production from 2000–2019 to 2040–2059 can be as high as +10% in South Texas or as low as −20% in Iowa. Larger turbines and moderate reductions in system costs can offset the largest projected decreases in wind resource, but much uncertainty remains in the extent to which wind resources will change and to what extent system costs can be reduced. An analysis of variance shows, in several states in the Midwest, uncertainty in future wind resources can influence the cost of wind energy nearly as much as uncertainty in future system costs.
ABSTRACT The western coast of the United States has abundant natural resources, including wind, sun, and water. Despite strong ocean winds, offshore wind energy (OSW) in the Western United States is in the early stages of technology development due largely to the deep water off the coast, which requires floating platforms rather than fixed‐bottom turbines. Floating OSW increases the technical difficulty of installation and thus the cost relative to both fixed‐bottom OSW and land‐based turbines. OSW is a potential generation option to help meet increasing demand on the west coast of the United States because it likely has fewer land‐use conflicts than other technologies and complements sources in the existing electricity supply by providing energy during times of high system stress. This paper examines the possible drivers and barriers to OSW deployment on the West Coast using the National Laboratory of the Rockies' state‐of‐the‐art capacity expansion model and the Regional Energy Deployment System (ReEDS) model. We use ReEDS to explore a multitude of future scenarios looking at key drivers for OSW deployment, including variations on the cost of OSW, electricity demand growth, and the availability of competing technologies to examine the factors that may play a role in OSW growth. Assuming coastal state policies such as renewable portfolio and clean energy standards remain in place, we find that OSW can play a role in meeting electricity demand and provide energy during stressful grid conditions and find that deployment from the least‐cost investment model ranges from 7.6 to 38 GW by 2045.
ABSTRACT Offshore wind farm operations depend on timely access to critical spare parts, yet maintenance planning is constrained by long procurement lead times, weather‐limited site access, and geographically distributed storage facilities. This study presents a practically oriented multiechelon spare parts inventory planning framework that supports engineering managers in making coordinated stocking decisions across distribution centers, local depots, and offshore sites. The proposed Spare Parts Inventory Policy (SPIP) integrates demand requirements, storage capacity, budget limitations, and service reliability targets into a unified planning model that remains transparent and practically interpretable for managerial use. Variability in component failures and offshore accessibility is incorporated through parameterized demand and service‐level inputs, allowing managers to evaluate trade‐offs without introducing unnecessary model complexity. Rather than introducing a new optimization algorithm, the contribution lies in translating offshore wind operational realities into a linear programming–based decision‐support framework that can be implemented with Python using the PuLP optimization package. A case study of a 60‐turbine offshore wind farm covering five classes of critical components demonstrates that the framework can reduce annual spare‐parts expenditures by approximately 27% relative to current industry planning practices while maintaining required service reliability. Scenario and sensitivity analyses further reveal how budget constraints, storage capacity, and service targets influence cost and inventory allocation decisions. The results provide actionable insights for maintenance planning, depot sizing, and inventory budgeting, offering a practical decision‐support approach for improving cost efficiency and maintenance readiness in offshore wind operations.
Optimizing wind farms is essential for designing efficient energy systems, especially as farms grow larger and span multiple sites. However, this optimization becomes increasingly challenging due to the rising computational cost associated with more turbines. Gradient-based optimization methods scale better than gradient-free approaches for large problems, but the most computationally expensive component remains the calculation of gradients for the objective function and constraint Jacobians. To address this, we propose leveraging sparsity to accelerate gradient evaluations and reduce the size of the constraint Jacobian. Wind farms naturally exhibit sparsity-many turbines do not influence each other under certain wind directions. However, unlike traditional sparse problems with fixed patterns, wind farm sparsity is dynamic, requiring new strategies to handle changing interactions efficiently. This paper presents a study of sparsity in wind farm optimization and introduces several methods to exploit it. These strategies are tested on multiple farms using the analytic Cumulative Curl model, with gradients computed via automatic differentiation (AD). The same sparsity-aware techniques are also applicable to finite difference (FD) methods, where they can yield even greater speedups due to the high cost of directional evaluations. Results show that sparse methods achieve up to a 10 speedup with less than 5% variance in optimized wake losses compared to traditional methods. These findings suggest that sparsity-aware optimization not only maintains solution quality but also scales efficiently with farm size, enabling more comprehensive design exploration at reduced computational cost.
ABSTRACT Physics‐based design optimization workflows thread the needle between computational cost limitations and simulation complexity, often compromising between modeling detail and the range of operating design conditions. Multipoint aerostructural optimization of wind turbine rotors has so far been confined to low‐fidelity analyses or to high‐fidelity studies with simplified structural models, leaving the most complex design trade‐offs unexplored. We close this gap by performing the first tightly coupled gradient‐based multipoint aerostructural rotor optimization using 3D aerodynamic and structural solvers with discrete coupled adjoints. The optimizer simultaneously varies blade planform, airfoil shapes, and structural thickness through more than 270 design variables, minimizing a weighted combination of rotor mass and power across multiple wind speeds. Applied to a modified DTU 10‐MW benchmark under conservative structural and aerodynamic constraints, our multipoint optimization reduces rotor mass by up to 36% and increases power by 12%–15% across the main operating conditions; biasing the objective toward power yields power gains up to 18% and a 17% mass reduction. For a nominal wind distribution, 3‐point rotor designs accounting for low RPM and high thrust conditions capture dominant trade‐offs and outperform single‐point designs. Adding two off‐design points changes individual‐condition power by less than 3% but leaves the weighted average within 0.5%, and the mass‐power bias has a stronger effect on the final design than the operating‐point weighting itself. Our framework extends naturally to richer load cases and site‐specific wind distributions, providing a basis for high‐fidelity multipoint design earlier in industrial workflows.
ABSTRACT The growing demand for offshore wind energy has led to a significant increase in wind turbine size and to the development of large‐scale wind farms, often comprising 100–150 turbines. However, the environmental impact of underwater noise emissions remains largely unaddressed. This paper quantifies, for the first time, the underwater aerodynamic noise footprint of three large offshore turbines (5, 10, and 22 MW) and wind farms composed of these turbines. We propose a novel methodology that integrates validated wind turbine noise generation (i.e., blade element momentum theory and Amiet) with plane wave propagation theory in different media, enabling turbine designers to predict underwater noise emissions. Our results indicate that the three turbines generate underwater noise levels that exceed the hearing thresholds of the low‐frequency hearing group in the range of 0.1–1 kHz. When scaled to represent wind farm conditions, the predicted noise levels may become detectable by additional hearing groups at frequencies up to 10 kHz. For the scenarios considered, the results suggest that aerodynamic noise from offshore wind farms could contribute to the underwater soundscape and may have implications for marine organisms. These findings highlight the importance of considering aerodynamic noise in environmental assessments of offshore wind energy development.
ABSTRACT This study develops an integrated GIS and multicriteria decision making (MCDM) approach supported by the Interval Type‐2 Fuzzy Risk Deviation Performance Index (IT2F‐RDPI) method to assess the wind energy potential in Diyarbakır Province. Fifteen criteria, including wind speed, slope, land use, elevation, soil structure, and proximity to transmission lines, were analyzed to identify suitable areas for wind power generation, resulting in approximately 7997.9 km2 of land being identified as suitable for wind energy production. The IT2F‐RDPI method evaluates the performance deviations and risks by considering uncertainties, providing a more reliable decision support framework. The results show that the Çüngüş and Kulp districts have the most favorable wind conditions with low risk and high energy potential, making them the most suitable areas for wind energy investment. Economic analyses confirm that wind energy investments in these areas are both technically and economically feasible. The findings highlight that the IT2F‐RDPI method offers a robust decision support tool for identifying the most suitable areas for wind energy investments.
Dynamic induction control (DIC), also known as pulse control, is a wake-mixing strategy intended to improve wind farm efficiency by mitigating wake-induced power losses. It typically employs harmonic collective blade pitching to excite wake instabilities, forming coherent pulsing structures that enhance turbulent mixing and accelerate wake recovery. Validation of this concept is primarily limited to fully waked conditions under fixed wind direction, while wind direction in the field is inherently variable. This study investigates the effectiveness of DIC in a closely spaced layout under varying inflow wind directions, considering both static and dynamic wake-impingement scenarios. To this end, wind tunnel experiments are conducted in a three-turbine configuration with a spacing of 2.5 , where denotes the rotor diameter. An open-loop controller is derived from experiments with fixed wind directions under uniform inflow, considering various combinations of front- and second-row turbine actuation with different pitch amplitudes and frequencies. Optimal pitch setpoints are stored in a look-up table and tested in a dynamic environment using a temporally scaled wind direction time series under both uniform and atmospheric boundary layer (ABL) inflow. Results from fixed wind direction experiments indicate that DIC provides power benefits within a wind direction range of , with optimal pitch setpoints remaining invariant for both actuated turbines. Specifically, sole front-row turbine actuation yields power gains of up to 5.7%, while adding second-row turbine actuation boosts these gains by up to 1.9%. Notably, under realistic wind direction variations, open-loop DIC demonstrates consistent wind farm power gains of up to 2.5% within the intended control range under both uniform and ABL inflow conditions. Overall, these findings highlight the effectiveness of DIC in a closely spaced layout, particularly at full or close to full wake-impingement conditions. Further research is needed to validate the identified wind direction range and the associated power uplift in wind farm layouts with larger turbine spacings.
ABSTRACT Optimising turbine layouts to maximise power output is crucial for wind farm development, particularly in complex terrain where analytical wake models fail to capture key flow physics. We present the first application of Bayesian optimisation (BO) combined with large eddy simulations (LES) for utility‐scale wind farm layout optimisation in synthetically generated, realistic and reproducible terrain. LES were performed using WInc3D, a fast, high‐fidelity simulator that resolves atmospheric boundary layers, terrain effects and turbine wake interactions. Initial tests show that standard black‐box BO achieves only marginal improvements in complex terrain, with limited capacity to identify high‐performing layouts. To overcome this, we developed physics‐informed sampling strategies that use terrain‐induced flow patterns from base‐flow LES that result in high‐performing initial layouts. We tested four 16‐turbine optimisation cases with different spacing constraints and wind directionality (single and multiple directions), evaluating 512 layouts with LES for each optimisation. Results show that optimal layouts exploit terrain features through clustering on elevated ridges for single‐direction cases, whereas multi‐directional optimisation requires balanced distributions to maintain performance. LES reveal complex flow phenomena, including terrain‐dependent wake behaviour, bidirectional turbine‐terrain interactions and non‐linear wake effects that necessitate high‐fidelity resolution. Analysis of the optimisation evolution reveals that BO serves primarily as a refinement tool, effectively exploiting physics‐informed initial designs but struggling to discover new optimal regions independently. These findings demonstrate that incorporating physical insight is essential for WFLO in complex terrain and that high‐fidelity simulations are necessary for reliable wind farm layout optimisation under such conditions.
