Wind Speed Estimation Research Articles

Comparative study of feature selection methods for wind speed estimation at ungauged locations

Wind speed estimation at ungauged locations is one of the preliminary steps for wind resource assessment. With the availability of high-resolution Digital Elevation Models (DEM) and remote sensing data, the number of potential wind speed predictors has grown substantially. The adequate spatial scale of these predictors is unknown a priori, leading to the use of multiple spatial scales of predictors in wind speed estimation models. Implementing a feature selection method as a pre-processing step of the analysis is necessary to avoid overfitting and the resulting potential model underperformance. This paper evaluated six feature selection methods (forward stepwise regression, Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net, maximum relevance minimum redundancy (MRMR), Genetic algorithm, and recursive feature elimination using support vector regression) for the estimation of different wind speed quantiles across Canada. The selected features were used to fit a regression-kriging model, and the importance of the predictors was evaluated with their associated regression coefficients. The results of the study showed that LASSO and MRMR are the most efficient algorithms with the least number of parameters to tune and good generalization performance. The study found that some predictors were more important for specific exceedance probabilities. The most important predictors were the distance from the coast and surface roughness length, regardless of exceedance probability.

Comparison of REWS and LIDAR-based feedforward control for fatigue load mitigation in wind turbines

While optimizing wind energy harvesting has always been the principal goal of wind turbine control, extensive research has been conducted on the use of advanced control methodologies aiming to maximize the turbine’s power generation while also achieving secondary performance objectives, such as fatigue load minimization. Some recent studies have investigated the application of a classical control methodology known as exact output regulation for improved wind turbine control. This control method is designed to achieve the rejection of known input disturbances, while also ensuring the system output tracks a desired reference signal. These studies all assumed wind preview information from LIDAR measurements was available to support the controller design.In this paper we investigate the use of exact output regulation for turbine control, using wind speed estimates obtained from rotor speed measurements. Simulated performance comparisons conducted over a wide range of mean wind speeds show that the exact output regulation methodology using wind speed estimates derived from rotor measurements can deliver similar reductions in fatigue loads to those obtainable with LIDAR measurements, without compromising power generation.

Quantifying Multifrequency Ocean Altimeter Wind Speed Error Due to Sea Surface Temperature and Resulting Impacts on Satellite Sea Level Measurements

Surface wind speed measurements from a satellite radar altimeter are used to adjust altimeter sea level measurements via sea state bias range correction. We focus here on previously neglected ocean radar backscatter and subsequent wind speed variations due to sea surface temperature (SST) change that may impact these sea level estimates. The expected error depends on the radar operating frequency and may be significant at the Ka band (36 GHz) frequency chosen for the new Surface Water and Ocean Topography (SWOT) satellite launched in December 2022. SWOT is expected to revolutionize oceanography by providing wide-swath Ka band observations and enhanced spatial resolution compared to conventional Ku band (14 GHz) altimetry. The change to the Ka band suggests a reconsideration of SST impact on wind and sea level estimates, and we investigate this in advance of SWOT using existing long-term Ku and Ka band satellite altimeter datasets. This study finds errors up to 1.5 m/s in wind speed estimation and 1.0 cm in sea level for AltiKa altimeter data. Future SWOT data analyses may require consideration of this dependence prior to using its radar backscatter data in its sea level estimation.

Predictive Control of a Wind Turbine Based on Neural Network-Based Wind Speed Estimation

Predictive control is an advanced control technique that performs well in various application domains. In this work, linearised control design models are first derived in state-space form from the full nonlinear model of the 5 MW Supergen (Sustainable Power Generation and Supply) exemplar wind turbine. Feedback model predictive controllers (FB-MPCs) and feedforward model predictive controllers (FF-MPCs) are subsequently designed based on these linearised models. A neural network (NN)-based wind speed estimation method is then employed to obtain the accurate wind estimation required for designing a FF-MPC. This method uses a LiDAR to be shared between multiple wind turbines in a cluster, i.e., one turbine is mounted with a LiDAR, and each of the remaining turbines from the cluster is provided with a NN-based estimator, which replaces the LiDAR, making the approach more economically viable. The resulting controllers are tested by application to the full nonlinear model (based on which the linearised models are derived). Moreover, the mismatch between the control design model and the simulation model (model–plant mismatch) allows the robustness of the controllers’ design to be tested. Simulations are carried out at varying wind speeds to evaluate the robustness of the controllers by applying them to a full nonlinear 5 MW Matlab/SIMULINK model of the same exemplar Supergen wind turbine. Improved torque/speed plane tracking is achieved with a FF-MPC compared to a FB-MPC. Simulation results further demonstrate that the control performance is enhanced in both the time and frequency domains without increasing the wind turbine’s control activity; that is, the controller’s gain crossover frequency (or bandwidth) remains within the acceptable range, which is about 1 rad/s.

Quantification and correction of motion influence for nacelle-based lidar systems on floating wind turbines

Abstract. Inflow wind field measurements from nacelle-based lidar systems offer great potential for different applications including turbine control, load validation, and power performance measurements. On floating wind turbines nacelle-based lidar measurements are affected by the dynamic behavior of the floating foundations. Therefore, the effects on lidar wind speed measurements induced by floater dynamics must be understood. In this work, we investigate the influence of floater motions on wind speed measurements from forward-looking nacelle-based lidar systems mounted on floating offshore wind turbines (FOWTs) and suggest approaches for correcting motion-induced effects. We use an analytical model, employing the guide for the expression of uncertainty in measurements (GUM) methodology and a numerical lidar simulation for the quantification of uncertainties. It is found that the uncertainty of lidar wind speed estimates is mainly caused by the fore–aft motion of the lidar, resulting from the pitch displacement of the floater. Therefore, the uncertainty is heavily dependent on the amplitude and the frequency of the pitch motion. The bias of 10 min mean wind speed estimates is mainly influenced by the mean pitch angle of the floater and the pitch amplitude. We correct motion-induced biases in time-averaged lidar wind speed measurements with a model-based approach, employing the developed analytical model for uncertainty and bias quantification. Testing of the approach with simulated dynamics from two different FOWT concepts shows good results with remaining mean errors below 0.1 m s−1. For the correction of motion-induced fluctuation in instantaneous measurements, we use a frequency filter to correct fluctuations caused by floater pitch motions for instantaneous measurements. The correction approach's performance depends on the pitch period and amplitude of the FOWT design.

Evaluation of Nakagami and Birnbaum-Saunders probability distribution for wind speed and power estimation

This study compares three probability density functions (PDFs) for understanding and estimating wind power based on wind patterns: the commonly-used Weibull distribution, the relatively-new Birnbaum-Saunders distribution, and the Nakagami distribution. The wind profile of Jumla, Nepal was analyzed using data from 2004 to 2014. The Nakagami distribution performed similarly to the Weibull distribution in terms of understanding wind patterns. However, for estimating wind power, the Nakagami distribution was found to be more effective than the Weibull distribution in most cases. The Birnbaum-Saunders distribution was found to be the least effective of the three PDFs compared.

DDM-Former: Transformer networks for GNSS reflectometry global ocean wind speed estimation

Global navigation satellite system reflectometry (GNSS-R) has shown a capability in recent years to be applied as a novel remote sensing technique to retrieve ocean wind speeds. The combination of GNSS-R observable delay-Doppler maps (DDMs) and deep learning algorithms provides the possibility to build an end-to-end pipeline for improving wind speed estimations. Recent studies have proven that data-driven approaches can be applied to generate enhanced estimation products. However, these are usually trained with convolutional neural networks (CNNs), which include inductive bias throughout the models. The inbuilt translation equivariance in CNNs represents an inexactitude for the feature extraction on DDMs. To address this issue, we propose Transformer-based models, named DDM-Former and DDM-Sequence-Former (DDM-Seq-Former), to exploit delay-Doppler correlation within and between DDMs, respectively. The advantages of our methods over conventional retrieval algorithms and other deep learning-based approaches are presented based on the Cyclone GNSS (CYGNSS) version 3.0 dataset. In addition, a comprehensive study on the attention mechanism for our models is demonstrated. The proposed DDM-Former yields the best overall performance with a root mean square error (RMSE) of 1.43m/s and a bias of −0.02m/s over the nine months test period. Moreover, with an RMSE of 2.89m/s and a bias of −1.88m/s, the proposed DDM-Seq-Former can promisingly improve the estimations in the cases with wind speeds higher than 12m/s. There are still opportunities for further enhancements in creating more robust models that could perform well in all wind regimes given a non-uniform wind distribution. We will make our code publicly available.

Influence of Atmospheric Stability on Wind Turbine Energy Production: A Case Study of the Coastal Region of Yucatan

Wind energy production mainly depends on atmospheric conditions. The atmospheric stability can be described through different parameters, such as wind shear, turbulence intensity, bulk Richardson number, and the Monin–Obukhov length. Although they are frequently used in micrometeorology and the wind industry, there is no standard comparison method. This study describes the atmospheric stability of a coastal region of Yucatan, Mexico, using these four parameters. They are calculated using six-month data from a meteorological mast and a marine buoy to determine atmospheric stability conditions and compare their results. The unstable atmospheric condition was predominant at the site, with an 80% occurrence during the measurement period, followed by 12% in neutral and 6% in stable conditions. Wind speed estimations were performed for each atmospheric stability scenario, and the variation in the energy produced was derived for each case. Unstable atmospheric conditions deliver up to 8% more power than stable conditions, while neutral conditions deliver up to 9% more energy than stable conditions. Therefore, considering a neutral state may lead to a considerably biased energy production estimation. Finally, an example calculation indicates that atmospheric stability is a crucial parameter in estimating wind energy production more accurately.

Nonlinear optimal control with effective wind speed estimation for maximum power extraction based on adaptive fuzzy logic controller and extended Kalman Filter

Nonlinear optimal control with effective wind speed estimation for maximum power extraction based on adaptive fuzzy logic controller and extended Kalman Filter

Wind and Turbulence Observations With the Mars Microphone on Perseverance

AbstractWe utilize SuperCam's Mars microphone to provide information on wind speed and turbulence at high frequencies on Mars. To do so, we first demonstrate the sensitivity of the microphone signal level to wind speed, yielding a power law dependence. We then show the relationship between the microphone signal level and pressure, air and ground temperatures. A calibration function is constructed using Gaussian process regression (a machine learning technique) taking the microphone signal and air temperature as inputs to produce an estimate of the wind speed. This provides a high rate wind speed estimate on Mars, with a sample every 0.01 s. As a result, we determine the fast fluctuations of the wind at Jezero crater which highlights the nature of wind gusts over the Martian day. To analyze the turbulent behavior of this wind speed estimate, we calculate its normalized standard deviation, known as gustiness. To characterize the behavior of this high frequency turbulent intensity at Jezero crater, correlations are shown between the evaluated gustiness statistic and pressure drop rates/sizes, temperature and energy fluxes. This has implications for future atmospheric models on Mars, taking into account turbulence at the finest scales.

A workflow for including atmospheric stability effects in wind resource and yield assessment and its evaluation against wind measurements and SCADA

Atmospheric stability conditions are known to impact the wind resource and yield assessments. However, they are too barely or not correctly taken into account in the industry due to several reasons such as limitations of commercial software or the relative inertia in updating industrial processes. This paper proposes a simplistic approach to improve wind resource and yield assessment certainty while keeping very similar software and industry processes. Two test cases are considered. First, the wind speed estimates made using a CFD software (Meteodyn 5.3) for different stability classes are compared to measurements obtained on a site with four met masts. Second, the wake losses obtained with a commercial yield assessment solver (WindPRO 3.5) considering different wake decay constant definitions are compared to SCADA data. In both cases, it appears that dividing the timeseries in “stability classes” and using corresponding stability parameters in the CFD and in the wake model parameters enable a reduction of the uncertainty. In the case of the energy estimation compared to SCADA, an improvement of 1.25% is obtained compared to the conventional approach.

A baseline for ensemble-based, time-resolved inflow reconstruction for a single turbine using large-eddy simulations and latent diffusion models

We are interested in reconstructing winds flowing through a turbine on a second-by-second basis over a 10 min window. Previously, we developed a machine learning algorithm that takes in a snapshot of wind speed measurements and generates ensembles of three-dimensional wind field estimates. Here, we use these estimates as initial conditions in large-eddy simulations and reconstruct atmospheric and turbine response dynamic quantities in a synthetic field campaign. In doing so, we establish a baseline for model validation that future time-aware data assimilation techniques will be compared to. In turbine-free case studies, ground truth wind speeds consistently fall within our estimated wind speed distribution for the first 100 s after the simulation start. In simulations with turbines, the wind estimates show a small bias of 0.10 m s−1 and good correlation of 0.80 during the first 100 s. During this window, our estimates of the Blade 1 bending moment and generator power typically span the ground truth, with the estimate of the former performing better overall. In summary, this approach shows promise as a stand-alone technique for reconstructing real-world inflow and turbine dynamics in 1−2 min windows and as a foundation for future time-aware data assimilation techniques.

A combined wind speed prediction model based on data processing, multi-objective optimization and machine learning

The ability to estimate wind speed accurately is crucial for optimizing the utilization of wind energy in power systems. The randomness of wind speed and the performance of prediction models are the two issues that have the greatest impact on wind speed prediction. This paper starts from the two aspects of digging wind speed characteristics and optimizing model performance. A new wind speed combined prediction model based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and improved artificial bee colony algorithm (IABC) to optimize least squares support vector machine (LSSVM) performance. In order to reduce the randomness of the wind speed sequence and increase its predictability, CEEMDAN is first used to break down the wind speed sequence. Secondly, an LSSVM prediction model optimized by an improved artificial bee colony algorithm is established for each subsequence obtained by decomposition; finally, the wind speed forecast value is created by superimposing the results of each subsequence’s prediction. Case studies demonstrate that the combined prediction model put out in this study outperforms other models in terms of prediction performance.

Uncertainty in the Mobile Observation of Wind

Air quality and greenhouse gas sampling from mobile platforms enables local to regional analyses of pollutant exposure, atmospheric chemistry, and emission sources. Simultaneous meteorological observations, particularly wind speed and direction, are often used to interpret measurements and construct emission fluxes. However, the wind arising from a moving platform contributes to the observed wind speed and direction, and this artifact requires adequate removal to best apply wind observations. Here, we calculate the theoretical limitations to the measurement of wind from a moving vehicle, assess the accompanying uncertainty, and apply these methods to an example transect across a plume of methane. The angle of the wind relative to the moving vehicle is a crucial determinant of the ability to distinguish a true wind and defines its uncertainty. Unlike a stationary wind measurement, the wind speed and direction contain complementary information that broadens the capability of the mobile anemometer. We find that the isolation of a true wind depends on the anemometer wind speed accuracy for true winds moving with or against the vehicle, while the anemometer directional accuracy is more important for crosswinds, such as is experienced when observing across a plume. The uncertainty in estimated wind speed has similar geometry, but the uncertainty in estimating true wind direction is the opposite: the accuracy of measured wind speed most greatly impacts crosswind direction. Exact values are determined by the specific accuracy limitations of the anemometer and vehicle speed, and the geometrical distributions vary. As a result, the characteristics of each mobile lab setup should be assessed individually to best inform meteorological analyses and observation route planning.

Implementation of hybrid wind speed prediction model based on different data mining and signal processing approaches

Accurate estimation of wind speed (WS) data, which greatly influences meteorological parameters, plays a vital role in the safe operation and optimization of the power system and water resource management. The study's main aim is to combine artificial intelligence and signal decomposition techniques to improve WS prediction accuracy. Feed-forward back propagation neural network (FFBNN), support vector machine (SVM) and Gaussian processes regression (GPR) models, discrete wavelet transform (DWT), and empirical mode decomposition (EMD) were used to forecast the WS values ​​1 month ahead in Burdur meteorology station. Statistical criteria such as Willmott's index of agreement, mean bias error, mean squared error, determination coefficient, Taylor diagram and regression analysis, and graphical indicators were used to evaluate the prediction success of the models. As a result of the study, it was determined that both wavelet transform and EMD signal processing increased the WS prediction performance of the stand-alone ML model. The best performance was obtained with the hybrid EMD-Matern 5/2 kernel GPR with test (R2:0.802) and validation (R2:0.606). The most successful model structure was obtained using input variables with a delay of up to 3 months. The study's results contribute to wind energy-related institutions in terms of practical use, planning, and management of wind energy.

Comprehensive validation of 68 wind speed models highlights the benefits of ensemble approaches

Many different wind speed models support assessing the wind climate. Depending on their development, spatiotemporal resolution, and intended use, typical wind speed model types are wind atlases, reanalysis products, regional climate, and global climate models. They realistically represent different properties of the wind climate. For the first time, 68 wind speed models of different types are assessed for their predictive power for (1) mean wind speed, (2) wind speed distribution, and (3) temporal variability. In addition, eight ensembles of individual models from each model type are created and validated. The hypothesis is tested that ensembles of wind speed models generally improve model accuracy. For the validation, 148 wind speed time series with a data availability of at least ten years in 1979 to 2014 were used. The measurement sites were separated into flat and open, and mountainous locations. The results reveal the improved model performance of ensembles compared to individual models. Only wind atlases are capable of estimating wind speed properties at mountainous locations. Their median coefficient of determination related to mean wind speed at mountainous sites is 0.81 but less than 0.15 for any other model type. One important conclusion is that individual models may not be appropriate for wind resource assessment without prior validation. Furthermore, ensembles are recommended to improve predictive power for any wind climate properties. The results indicate how well the different model types can reproduce the analyzed wind speed characteristics.

Assessment of wind speeds along the damage path of the Alonsa, Manitoba EF4 tornado on 3 August 2018

Given the impracticality of attempting to directly measure wind speeds in tornadoes, wind speed estimation typically relies on the assessment of damage to structures and vegetation using classifications described in the Enhanced Fujita (EF) Scale. The advent of technology enabling the collection of large amounts of data, including detailed ground, drone, and aerial imagery, has led to a growth in research on non-conventional approaches for estimating tornado wind speeds. Research methods focused on defining the tornadic wind field based on forensic analysis of damage observations have shown promise for improving tornado assessments in a quantitative manner. In this study, novel methods for collecting forensic data following tornadoes are presented. Data from the Alonsa, MB tornado are applied to estimating the wind field along the damage path using treefall pattern analysis and threshold debris flight speed calculations. Comparison of the resulting wind speed estimates show reasonable agreement, with maximum speeds from both methods in the EF5 range. These research methods yield higher wind speeds than the maximum value obtained from the conventional EF-Scale assessment, which is in the low-end of the EF4 range based on a wood-frame house with sub-standard construction that was swept entirely from its foundation. Further work is still needed to make these methods operational for routine tornado intensity estimates.

Estimating hub-height wind speed based on a machine learning algorithm: implications for wind energy assessment

Abstract. Accurate estimation of wind speed at wind turbine hub height is of significance for wind energy assessment and exploitation. Nevertheless, the traditional power law method (PLM) generally estimates the hub-height wind speed by assuming a constant exponent between surface and hub-height wind speed. This inevitably leads to significant uncertainties in estimating the wind speed profile especially under unstable conditions. To minimize the uncertainties, we here use a machine learning algorithm known as random forest (RF) to estimate the wind speed at hub heights such as at 120 m (WS120), 160 m (WS160), and 200 m (WS200). These heights go beyond the traditional wind mast limit of 100–120 m. The radar wind profiler and surface synoptic observations at the Qingdao station from May 2018 to August 2020 are used as key inputs to develop the RF model. A deep analysis of the RF model construction has been performed to ensure its applicability. Afterwards, the RF model and the PLM model are used to retrieve WS120, WS160, and WS200. The comparison analyses from both RF and PLM models are performed against radiosonde wind measurements. At 120 m, the RF model shows a relatively higher correlation coefficient R of 0.93 and a smaller RMSE of 1.09 m s−1, compared with the R of 0.89 and RMSE of 1.50 m s−1 for the PLM. Notably, the metrics used to determine the performance of the model decline sharply with height for the PLM model, as opposed to the stable variation for the RF model. This suggests the RF model exhibits advantages over the traditional PLM model. This is because the RF model considers well the factors such as surface friction and heat transfer. The diurnal and seasonal variations in WS120, WS160, and WS200 from RF are then analyzed. The hourly WS120 is large during daytime from 09:00 to 16:00 local solar time (LST) and reach a peak at 14:00 LST. The seasonal WS120 is large in spring and winter and is low in summer and autumn. The diurnal and seasonal variations in WS160 and WS200 are similar to those of WS120. Finally, we investigated the absolute percentage error (APE) of wind power density between the RF and PLM models at different heights. In the vertical direction, the APE is gradually increased as the height increases. Overall, the PLM algorithm has some limitations in estimating wind speed at hub height. The RF model, which combines more observations or auxiliary data, is more suitable for the hub-height wind speed estimation. These findings obtained here have great implications for development and utilization in the wind energy industry in the future.

High-altitude vertical wind profile estimation using multirotor vehicles.

Capturing vertical profiles of the atmosphere and measuring wind conditions can be of significant value for weather forecasting and pollution monitoring however, collecting such data can be limited by current approaches using balloon-based radiosondes and expensive ground-based sensors. Multirotor vehicles can be significantly affected by the local wind conditions, and due to their under-actuated nature, the response to the flow is visible in the changes in the orientation. From these changes in orientation, wind speed and direction estimates can be determined, allowing accurate estimation with no additional sensors. In this work, we expand on and improve this method of wind speed and direction estimation and incorporate corrections for climbing flight to improve estimation during vertical profiling. These corrections were validated against sonic anemometer data before being used to gather vertical profiles of the wind conditions around Volcan De Fuego in Guatemala up to altitudes of 3000m Above Ground Level (AGL). From the results of this work, we show we can improve the accuracy of multirotor wind estimation in vertical profiling through our improved model and some of the practical limitations of radiosondes that can be overcome through the use of UAS in this application.

Martian Wind and turbulence heard by the SuperCam microphone on the perseverance rover

On top of listening to laser shots, rover sounds and the Ingenuity rotorcraft, SuperCam’s Mars microphone has recorded over 7 hours of ambient background noise on Mars. These background recordings contain signal due to the Martian wind. Through a comparison to the meteorological data recorded by the MEDA (Mars Environmental Dynamics Analyzer), we can determine the relationships between the microphone data, the wind and the atmospheric stability. Based on these relationships, we have determined a way to estimate the wind speed using the microphone through Gaussian process regression, a machine learning technique. Owing to the sampling rate of 25 000 samples per second, the microphone data can be used to examine Mars’ atmospheric dynamics at high frequencies, as yet unexplored on Mars. We will demonstrate how the wind speed estimates from the microphone provide an assessment of turbulence at fine scales, shedding light on the dissipative regime on Mars. One particularly interesting signal recorded by the microphone was a dust devil, which had fast varying winds within the walls of its vortex and signal from dust particles hitting the rover. Combining the microphone data with information from the MEDA sensors and navigation camera (Navcam) images enabled a full parameterization of this event.