Distributional shifts across wind farms and non-stationary dynamics pose significant challenges to accurate and generalizable wind power forecasting. This paper proposes UDA-MTVIB, an unsupervised domain-adaptive multi-site temporal variational information bottleneck framework. The model integrates a dynamic-gated LSTM encoder with uncertainty-aware gating to regulate temporal memory updates under domain shift. A dual variational information bottleneck is employed to (i) compress task-relevant temporal representations and (ii) regularize domain-related latent factors, which are aligned via adversarial training for robust domain alignment. Unsupervised domain adaptation is achieved by adversarial alignment between labeled source data and unlabeled target-farm inputs, without target power labels. Experiments on WIND Toolkit data across seven U.S. wind farms demonstrate that UDA-MTVIB achieves strong robustness and generalization under cross-farm and cross-year shifts, improving predictive accuracy and stability across multiple metrics. This work provides an information-theoretic perspective for domain-adaptive temporal forecasting in renewable energy systems.
This study proposes a flexible power management strategy (FPMS) for a hybrid renewable energy system integrating a doubly fed induction generator-based wind turbine (DFIG-WT) and a photovoltaic generator (PVG). The system employs three power electronic converters to regulate power flows and interface the generators with the electrical grid. A key innovation of this work is the integration of the proposed FPMS with a direct connection of the PVG to the DC-link of the DFIG-WT, enabling a single grid-side converter (GSC) for grid coupling, thereby reducing switching losses and overall system costs. The FPMS dynamically manages power distribution, ensures uninterrupted supply to both DC and AC loads, and coordinates the rotor-side converter (RSC) and GSC according to sub- and super-synchronous DFIG operating modes. Five operational modes are defined to optimize DC/AC and AC/DC power flows, enhancing system flexibility and maximizing renewable energy utilization. Simulation results demonstrate that the implemented FPMS reduces grid dependency by up to 82.7% for DC loads and 4.76% for AC loads, increases energy injection into the grid, and maintains stable operation under variable wind, solar, and load conditions. These findings confirm that the suggested FPMS improves energy self-sufficiency and operational flexibility, providing an effective energy management framework for the integration of distributed renewable energy sources into the grid.
This paper provides a comprehensive analytical approach based on steady state and transient analysis of turbine power system using either a brushless DC generator or permanent magnet generator with Sinusoidal Waves-Form. The added value of this study is the use of a steady-state analytical model enabling the design and optimization of the power chain manufacturing parameters, given that the use of the steady-state analytical modeling technique is highly compatible with high-dimensional stochastic optimization algorithms. For the same turbine and generator stator structures, the energy recovered from the turbine with permanent magnet generator with Sinusoidal Waves-Form is compensating by an increase of the generator angular speed, since the electromagnetic torque acting as a brake is greater in the case of the brushless DC generator. The global model of the power chain is implemented under the Matlab-Simulink simulation environment. This model is validated against the classic model of the wind turbine using the SimPowerSystem library of power component models integrated under the Matlab-Simulink simulation environment.
Rigid one-mass drivetrain models simplify maximum power point tracking (MPPT) controller synthesis, but their use can create a model-controller mismatch when the implemented wind turbine has a flexible shaft. This study evaluates that mismatch using a nonlinear MATLAB/Simulink model of a 1.5 MW variable-speed horizontal-axis wind turbine. Three PI-based cases are compared: (A) a controller tuned with a one-mass model and applied to a two-mass plant, (B) a controller tuned with the two-mass plant, and (C) the matched controller supplemented by active torsional damping. The gains are selected from linearized one-mass and two-mass models using pole-placement and frequency-domain constraints, while identical saturation and anti-windup settings are retained. The controllers are tested for 60 s under a reproducible wind profile with a mean speed of 11.6 m/s, deterministic low- and medium-frequency components, a gust, and low-intensity random turbulence (sigma = 0.20 m/s; turbulence intensity approximately 1.7%). Under these specific conditions, the three cases yield 0.004125 MWh, a mean absolute generator-speed tracking error of 2.14 rad/s, a tracking-error standard deviation of 4.36 rad/s, and a maximum low-speed-shaft torque of 10.924 MN m. No amplified peak is observed near the calculated torsional frequency. The contribution is therefore a conditional counterexample and a reproducible benchmark showing that rigid-model tuning is not necessarily destabilizing for this parameter set. It is not a general proof of safety: the result is limited to the tested turbine, PI architecture, short simulation horizon, high inherent damping, and low turbulence level.
Wind energy represents one of the important sources of renewable energy (RE) and plays a pivotal role in the international decarbonization of energy systems. The novelty of the research work lies in the development and evaluation of a multi-horizon forecasting strategy under sparse-data conditions, where limited but evidently meaningful meteorological variables, that is, wind speed, wind direction, temperature, air pressure, and humidity, are used for short-term prediction, because these variables are important for accurate forecasting. Numerous Machine Learning (ML) and Deep Learning (DL) models, including Decision Tree (DT), Random Forest (RF), Support Vector Regression (SVR),Extreme Gradient Boosting (XGB), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU), the models were developed, and their performance was compared and analyzed for 10-minute, 30-minute, and 60-minute onward wind speed forecasting, and the results show that the Random Forest model achieved strong short-term forecasting performance at the 10-minute horizon with Mean absolute error (MAE) of 0.4364, Mean square error (MSE) of 0.3544, and Coefficient of determinations (R 2 ) of 0.9157, while the GRU model demonstrated modest temporal learning capability with R 2 values beyond 0.91 for short-term prediction. At the 30-minute horizon, GRU achieved the highest R 2 of 0.877, whereas RF maintained low prediction errors with MAE of 0.5628 and MSE of 0.5756, and for the 60-minute horizon, RF remained the most stable model, producing MAE of 0.6938, MSE of 0.8738, and R 2 of 0.7778, indicating better robustness than DL models under limited data conditions. The results validate that ensemble learning models are more reliable for sparse Meteorological datasets, whereas DL models are effective for capturing short-term temporal patterns, subsequently this research provides a practical forecasting framework for wind farms, offering improved operational planning, reduced forecasting uncertainty.
This study presents a manufacturability-oriented aerodynamic design workflow for a 10 kW horizontal-axis wind turbine blade. Thirty-five NACA-series airfoils were screened using the Robust Envelope Minimax with Smoothness (REMUS) framework, which ranks candidates according to regret, robustness, drag-bucket width, and stall behavior. NACA 4518 was selected for the modeled aerodynamic span. A nonlinear reference geometry was then obtained using blade element momentum theory with Prandtl tip-loss and high-induction corrections. The chord and twist distributions were linearized to generate 225 prescribed candidate blades. The best candidate within this finite set achieved a design-point power coefficient of 0.4707, compared with 0.4516 for the corrected-BEM baseline. However, this improvement was confined to a narrow region around tip-speed ratios of 7-8, and the linearized candidate performed worse under off-design conditions. A numerical cross-check against QBlade produced differences below 3.2%. Because both implementations rely primarily on BEM-type aerodynamic modeling, this agreement is interpreted as an implementation-level consistency check rather than independent experimental or high-fidelity validation.
Accurate wind speed estimation is required for integrating wind energy into Cameroon’s hydropower-dependent national grid. However, Cameroon’s diverse climatic zones present significant challenges for modelling. This study develops and evaluates a Hybrid Improved Genetic Algorithm-Support Vector Machine (IGA-SVM) framework for wind speed prediction across five climatic regions, with subsequent conversion to wind power density (WPD). Using 40 years of NASA’ Prediction of Worldwide Energy Resources (POWER) meteorological data (temperature, relative humidity, atmospheric pressure, rainfall, wind direction, snowfall, wind speed, and snow depth), the Hybrid IGA-SVM eliminates manual hyperparameter tuning by optimizing the SVM’s box constraint (C), kernel scale (γ), and epsilon-insensitivity (ε), independently for each zone. The model achieves R 2 values from 0.7956 to 0.8634 across all terrains. The Adamawa Plateau shows the highest stability and predictability, while the Sudano-Sahelian zone shows the greatest wind power density despite seasonal volatility. However, none of the zones reached the National Renewable Energy Laboratory (NREL) Class 3 thresholds (300–400 W/m 2 ); the highest wind power density (WPD) (81 W/m 2 in the Western Highlands) falls into Class 2, indicating suitability for small to medium scale or hybrid applications. The proposed model can serve as a preliminary screening tool for regional energy planning in Cameroon. All recommendations require on-site validation and economic analysis (e.g., levelized cost of energy (LCOE)) before investment decisions. Importantly, all zone-specific recommendations in this paper are based exclusively on technical metrics (R 2 , Root Mean Square Error (RMSE), wind speed, and wind power density); no economic analysis (levelized cost of energy, net present value, or payback period) were performed, and thus these recommendations should be interpreted as pre-feasibility guidance for prioritizing zones for further study, and not as final investment advice.
In developing nations like Egypt, a solar PV-based water-pumping facility is affordable. A comparison of different simulation software tools for renewable energy systems is presented here. This study examines remote solar water pumping systems in the climate-sensitive Egyptian region, accounting for variables such as site location, system size, and performance. The remote location in southern Egypt (Fayoum town) provides the data inputs. When designing and evaluating the performance of a solar water pumping system, technical, economic, and environmental factors are considered. Homer (Hybrid Optimization of Multiple Electric Renewables) Pro and iHoga (enhanced hybrid optimization genetic algorithm) software are used to build and simulate the power system required to meet this energy demand. The monitoring system’s collected data indicates that the highest daily water volume was 726 m 3 . Based on solar yield, pump efficiency, energy losses, and performance ratio (PR), a comparison of the four regions is carried out. Solar water pumping system (SWPS) employing HOMER has an average monthly electric output power of roughly 0.6 MWh. For HOMER and iHoga, energy production is around 2000 kWh and 1800 kWh, respectively, while excess energy is 600 kWh and 500 kWh, respectively. The analysis cost of the suggested SWPs is split as follows: 65% for batteries, 10.5% for inverters, and 24.5% for photovoltaic systems. For HOMER, the net present cost is 52,000 USD, and for I Hoga, it is 72,000 USD. For HOMER and iHOGA, the energy costs are 0.185 and 0.221 per kWh, respectively. According to two simulation analyses, the performance ratio ranges from 59% to 70%, and the overall pump efficiency ranges from 57% to 66%. From an environmental perspective, this proposed SWPS avoided 250 kg of CO 2 annually when using iHoga and 300 kg of CO2 annually when using HOMER. HOMER is the most dependable and effective simulation software for this proposed SWPS, and the results help make policy decisions on where to locate the optimal solar pumping station based on climate conditions, especially in rural settlements.
Accurate wind speed downscaling is essential for meteorological applications and reliable station-scale wind estimation. This study presents a systematic comparison of recurrent deep learning architectures, including RNN, GRU, LSTM, Bidirectional LSTM, Sequence-to-Sequence LSTM, and Stacked LSTM, against conventional statistical models for wind speed downscaling using paired IMD station observations and NASA MERRA-2 reanalysis data for Pune, India. All models were evaluated under identical preprocessing and training protocols to ensure a fair comparison. Among the standard recurrent architectures, the Stacked LSTM achieved the best predictive performance. An Attention-Enhanced LSTM was subsequently evaluated as an extension to investigate the benefits of incorporating a soft-attention mechanism. It achieved the highest overall accuracy, with an RMSE of 0.5034 and an MAE of 0.3650, demonstrating additional performance gains over the standard recurrent architectures. The findings provide a robust benchmark for deep learning-based wind speed downscaling and future atmospheric prediction studies.
Modelling wind speed at hourly resolution over multi-year horizons remains challenging, particularly in tropical regions influenced by strong climatic variability. This study presents a hybrid Fourier–PSO framework for four-year ahead wind speed forecasting using ERA5 reanalysis data for Barranquilla, Colombia. The methodology combines spectral decomposition through the Fourier Transform with a signal-to-noise ratio criterion to identify dominant harmonic components. A reduced subset of harmonics is subsequently adjusted using Particle Swarm Optimization, while keeping the original frequencies fixed to preserve physical consistency. The model is trained using 38 years of historical data and evaluated over an independent four-year test period. The hybrid model achieved Root Mean Square Error values close to 2.0 m/s, Mean Absolute Error values between 1.6 and 1.7 m/s, and correlation coefficients ranging from 0.65 to 0.70 over the independent test horizon. The results support a stable and interpretable approach for long-term wind resource characterization in tropical environments.
This study provides a decision-grade synthesis of floating vertical axis wind turbines (F-VAWTs) for deep-water wind energy, integrating device physics, platform dynamics, farm effects, and bankability over 2012–2025. Standardized searches across Scopus, Web of Science, Google Scholar, ScienceDirect, Taylor & Francis, and SpringerLink returned 237 records; a PRISMA-aligned, motion-aware eligibility cascade and expert adjudication refined this to 123 critical studies. Publication activity rose from one paper in 2012 to a peak of 18 in 2022, then stabilized at 14 in 2023 and 12 per year in 2024–2025, signaling maturation toward fully coupled aero-hydro-servo-elastic modeling. The corpus is dominated by engineering (83) and energy (76) outputs, with growing emphasis on environmental science (42) and mathematics (27), while computer science (5) and earth & planetary sciences (4) remain under-represented—highlighting opportunities for digital twins, physics-informed ML surrogates, and uncertainty-aware control. Geographically, China leads (53 studies), followed by the United Kingdom (20), with Norway and Italy (14 each) forming a robust second tier. Key outlets were Energy and Renewable Energy (∼13 each), Ocean Engineering (12), and Energies (11). Synthesized technical results show that controller-in-the-loop co-design with floater/mooring stiffness can mitigate negative aero-damping and reduce fatigue-critical load spectra; motion-conditioned array layouts (including counter-rotation and staggered spacing) can accelerate wake recovery while easing cable/tendon demands; and validated digital-twin workflows can shorten design loops and enable condition-based O&M. A phase-gated global roadmap is proposed: standards and pilots (2025–2030), 100–300 MW arrays (2031–2040), and autonomous, resilience-oriented fleets (2041–2050), supported by finance and permitting instruments that de-risk motion-coupled offshore assets.
This study provides a comprehensive assessment of Brazil’s wind energy potential. The analysis encompasses electricity generation capacity and wind resource availability across multiple heights, covering both onshore and offshore contexts. In addition, the study examines the country’s electricity consumption profile and the broader generation–demand landscape, emphasizing the strategic role of wind power within the Brazilian energy matrix. A comparative perspective is also incorporated, positioning Brazil relative to other Latin American countries and to the BRICS group. The findings highlight the rapid and sustained expansion of Brazil’s wind sector, which is consolidating itself as one of the country’s most promising renewable energy sources. This evolution underscores wind energy’s growing contribution to diversifying the electricity mix, supporting a higher share of renewable generation. The study further reinforces the importance of policy frameworks that prioritize investment in wind energy as a means to strengthen energy security, improve competitiveness, and promote long-term sustainability.
This study presents a unified modeling framework for unconventional horizontal axis wind turbines (HAWTs) that utilize axially extended, non-radial blade geometries. Recent bio-inspired and helical designs have reported performance levels that are difficult to interpret within the standard single-plane actuator-disc framework: they overcome the classical Betz limit which is derived for an ideal single rotor operating under one-dimensional axial-flow assumptions. To address this gap in the technology description, we introduce the Generalized Multi-Stage Actuator Disc Model, which discretizes the rotor into a sequence of interacting aerodynamic stages aligned with the flow axis. We rigorously evaluate four topological configurations, Solid Divergent, Hollow Divergent, Solid Convergent, and Hollow Convergent, using numerical optimization to determine their theoretical performance limits. Our results show that all multi-stage configurations converge to a maximum theoretical efficiency of η ≈ 2 / 3 (approximately 66.7%), exceeding the single-stage Betz limit of 16 / 27 (59.3%), while remaining consistent with tandem-disc extensions of momentum theory. Furthermore, the analysis reveals that while constant-radius geometries are optimal under standard area normalization, divergent geometries evaluated under mass-conserving conditions behave as ideal diffusers, theoretically capable of recovering nearly all wake energy. We also quantify the impact of rotor solidity, showing that hollow configurations require maximized annular thickness to fully exploit the multi-stage effect. These findings provide a physical interpretation of non-conventional horizontal axis wind turbines, as geometries that lie outside the classical single-disc assumptions and establish a theoretical framework for designing turbines suitable for low-wind-speed environments.
As wind energy capacity surpasses 1136 GW globally, ML technologies prove essential for achieving renewable energy targets and grid stability requirements. Machine learning (ML) has emerged as a transformative technology for wind energy systems, revolutionizing forecasting accuracy, operational efficiency, and system reliability. This comprehensive review synthesizes recent advances across 500+ peer-reviewed studies from 2020 to 2025, revealing 15–40% performance improvements over traditional methods across all major applications. Deep learning approaches achieve up to 99% accuracy in fault detection while optimized forecasting systems reduce Mean Absolute Percentage Error (MAPE) to 5–12% and demonstrate 20% increases in energy value through advanced prediction capabilities. The review examines ML applications spanning wind power forecasting, turbine control optimization, predictive maintenance, and emerging technologies including digital twins and physics-informed neural networks. Critical challenges including data availability, model interpretability, and cross-site generalization are addressed, while future research directions emphasize physics-informed ML, federated learning, and explainable AI approaches.
With the fast developments of power electronics, shunt Flexible AC Transmission Systems (FACTS) devices, particularly Static Synchronous Compensators (STATCOMs), have been proposed and implemented in many wind energy systems (WESs). STATCOMs are used to enhance the integration of WESs into existing power systems by controlling power flow between the WESs and the grid, thereby ensuring compliance with grid codes. These devices enhance the grid-connected WES stability and reliability against different fault scenarios. This paper provides a comprehensive review of the application of STATCOMs along with their control methodologies. The review includes the role and configuration of the STATCOM in WES, as well as its role in mitigating various fault events, including three-phase to ground faults, sag-swell conditions, and ferroresonance events. The paper also discusses the control methodologies in STATCOM-WES, including classical PI controllers, optimized PI controllers, and adaptive control techniques, and their performance against different fault scenarios. Furthermore, the future trends in shunting FACTS technology for renewable energy integration are also explored. The review combines results from numerous case studies and simulation-based assessments, highlighting the role of STATCOM in WES to enhance system performance while supporting the low-voltage ride-through (LVRT) and fault ride-through (FRT) capabilities of grid-tied WESs. The study finds that STATCOMs are necessary for the reliable integration of WES into the power grid. It also highlights future trends, such as hybrid systems with energy storage and wide-bandgap semiconductors, that will enhance their efficiency.
This paper provides a comparative analysis between two turbine structures that differ only in the use of either a structure of permanent magnets generator with sinusoidal waveforms or its equivalent structure of permanent magnets generator with trapezoidal waveform. The design parameters of turbines using these two structures of generator are optimized by a developed Genetic Algorithm, in the goal to maximize the recovered power. Indeed, the comparison between the two wind turbines leads to the choice of the turbine using a trapezoidal waveform generator enabling an approximately 25% increase in recovered energy. This result represents the principal finding of this study. In addition, this study is completed by the analysis of the energy performance of the chosen wind turbine by the use of a booster chopper for recovered energy optimization. A braking system regulating the generator speed is added to these turbine structures. The braking system plays an important role in the protection of the turbine against over-currents and subsequently against the burning of the turbine considering, that the magnitude of the back electromotive forces increases depending on the elevation of the angular speed of the turbine. The overall study is validated by implementation of the studied models under SIMULINK simulation environment.
Among the various alternatives to traditional generators, the brushless doubly fed induction machine (BDFIM) is particularly well suited for variable-speed applications. However, advanced control strategies are required to fully exploit its capabilities. This study therefore proposes a novel cascaded field-oriented control (FOC) strategy incorporating complex fuzzy logic controllers (CFLCs). The primary objective of this approach is to introduce complex-valued control while reducing the overall dimensionality of the control scheme by exploiting both the magnitude and phase components of the complex control parameters. To assess the effectiveness of the proposed CFLC structure, simulations were carried out in the MATLAB/SIMULINK environment on a BDFIM system combined with a Maximum Power Point Tracking (MPPT) algorithm for wind energy generation. The results demonstrate clear and significant advantages over a standard Complex Proportional Integral (CPI) controller. The CFLC achieves a more efficient trajectory in the complex plane, reflecting smoother and more direct transitions. Additionally, it exhibits reduced power coupling and overshoot, an accelerated phase response, and enhanced voltage regulation, which may mitigate equipment stress. These findings underscore the effectiveness of the CFLC and highlight promising alternatives, thereby providing a strong impetus for continued research in the emerging domain of complex fuzzy control.
The pre-feasibility analysis of a commercial wind turbine occupies a long-term testing duration ranging between two to 4 years. This exercise specifically requires accurate knowledge of wind speed for the tested location with its applications also in informed decision-making and optimization. Though approaches for estimation of wind speed exist, they remain truly unvalidated using larger sample of data for a non-trained location. Furthermore, consideration of ground-based true measurements for development of estimator are also a concerning fact. So, the objective of presented investigation is to formulate a short-term data driven model for day ahead prediction of wind speed covering wide range of latitudes from 14.27 0N 74.44 0E to 17.32 0N 76.83 0E. The model is trained with actual set of data obtained from ground-based measurement stations set up by IMD (Pune). Six different individualistic and ensemble hybrid models are formulated to accurately forecast day ahead wind speed applicable for regions in Karnataka, India. Developed individualistic and hybrid models are derived from the statistical and machine learning approaches like seasonal ARIMA (SARIMAX), ANN, and LSTM. Among the developed models, LSTM and ANN-based models resulted in better forecasting accuracy than SARIMAX and the hybrid models possessing a mean absolute error of 0.133 and 0.068, respectively, for validation dataset 1. The MSE for the same are 0.038 sq. km/h and 0.009 sq. km/h, respectively. Similarly, the value of MSE for proposed ANN and LSTM for validation dataset 2 are 0.034 sq. km/h and 0.033 sq. km/h, respectively, and they possess a MAE of 0.11 km/h and 0.0095 km/h, respectively. In addition, statistical significance testing using the Diebold-Mariano test and residual diagnostics were conducted to verify the robustness of the forecasting results. The formulated models are also compared with the reported literature-based models and as a result of comparison the proposed models reflect better accuracy. By leveraging both ground measurements and theoretical understanding, the proposed short-term estimator for wind speed seeks to improve the accuracy and can be employed for prediction of wind speed.
Wind turbine towers (WTT), as tall slender structures, are exposed to complex loading conditions that may cause excessive oscillations affecting structural fatigue life and cost of maintenance; in this regard, vortex-induced vibration (VIV) is a challenging phenomenon to mathematically formulate and attenuating its resulting response in structures through added structural damping requires certain measures especially during assembly. Passive structural control devices have proven to be an effective solution in reducing structural vibrations and certain attributes of novel devices proposed in the literature could be exploited in new applications in a manner to overcome the limitations imposed by traditional dampers. In this paper, the effect of a magnetic damper with nonlinear stiffness and eddy current damping on VIV in off-shore WTT at four stages of installation is numerically investigated. Maximum VIV response of the tower is formulated using a spectral model for the off-shore DTU 10 MW reference wind turbine (RWT), taking into account the feasibility of implementing the device in terms of required space. The study reveals that the nonlinear damping device, in a compact and portable configuration, is capable of mitigating relative maximum normalized displacement and harmonic response due to VIV in off-shore WTTs at different installation stages. Based on the obtained results, the damper may be used as an installation aiding device to help facilitate tower erection and turbine assembly.
