Wind Speed Datasets Research Articles

A Hybrid Intelligent Algorithm for Deterministic and Probabilistic Interval Predictions of a High-Speed Railway Wind Warning System

A wind warning system (WWS) is designed to ensure the operational safety of trains amid various wind environments. The system integrates the Internet, cloud computing, and virtual instrumentation technologies to automatically collect wind speed data and intelligently output alert information. To precisely predict the wind speed, a hybrid intelligent algorithm termed VMD–SSA–GRU comprising variational mode decomposition (VMD), sparrow search algorithm (SSA), and gated recurrent unit (GRU) is proposed and integrated into the WWS. Besides, the quantile regression (QR) is applied to wind speed uncertainty estimation, and the probability density functions of partial intervals are further evaluated by an improved kernel density estimation method. Two typical wind speed datasets (typhoon and normal wind conditions) measured by the WWS are used for validation of the applicability of the developed forecasted model. Through comparison to seven single and four combined models, the developed model presents the highest forecast accuracy in both definitive and uncertainty predictions even in typhoon wind conditions. The study demonstrates that the WWS integrated with the proposed hybrid intelligent algorithm can provide sensible warning for the safety of train operation.

Wind speed prediction model based on multiscale temporal‐preserving embedding broad learning system

AbstractThe inherent randomness and intermittent nature of wind speed fluctuations pose significant challenges in accurately predicting future wind speeds. To address this complexity, a wind speed prediction model based on a multiscale temporal‐preserving embedding broad learning system (MTPE‐BLS) is proposed. MTPE‐BLS used the localised behaviour of wind speed data, which is simpler to model and analyse than global patterns. Firstly, frequency clustering‐based variational mode decomposition (FC‐VMD) is proposed to deal with the non‐stationary wind speed data into multiple intrinsic mode functions (IMFs). Then, temporal‐preserving embedding (TPE) is proposed to extract the underlying temporal manifold structure from the decomposed IMFs. Finally, the extracted features are mapped into the broad learning system (BLS) to establish an accurate prediction model. Experimental results on two real‐world wind speed datasets demonstrate the best performance of the proposed MTPE‐BLS model compared to that of others. Compared to the original BLS, the MTPE‐BLS achieves significant improvements, reducing the root mean square error (RMSE) and mean absolute error (MAE) by an average of 48.57% and 47.72%, respectively.

STGPT2UGAN: Spatio-Temporal GPT-2 United Generative Adversarial Network for Wind Speed Prediction in Turbine Network

In a turbine network, an accurate wind speed prediction plays a key role in improving the efficiency of turbine operations. At present, Large Language Models (LLMs) have shown their strong capabilities in time sequence forecasting. Unfortunately, the existing LLM-based models face the following two difficulties in precise wind speed prediction: (1) These models fail to capture spatio-temporal correlations. (2) These models rely exclusively on supervised learning to train. Supervised learning updates parameters by optimizing a loss function, which focuses solely on improving accuracy on a specific regression or classification task rather than understanding the true data distribution. To tackle these difficulties, we devise a novel wind speed prediction architecture entitled Spatio-Temporal GPT-2 United Generative Adversarial Network (STGPT2UGAN). In particular, in order to model spatio-temporal correlations, we first design a Spatio-Temporal GPT-2 (STGPT-2), which comprises a Spatio-Temporal Block and the variant of GPT-2. Second, to improve the training of STGPT-2, a Generative Adversarial Network (GAN) is introduced to develop an adversarial training strategy. The strategy combines unsupervised training and supervised training in a complementary fashion. The unsupervised training promotes STGPT-2 to learn the distribution of true data through minimizing the adversarial loss. The supervised training intends to align the true values and the predicted values. We conduct the experiments on one real-world wind speed dataset. The experimental results verify that STGPT2UGAN outperforms the state-of-the-art benchmarks in terms of prediction precision.

Ultra-Short-Term Wind Power Forecasting in Complex Terrain: A Physics-Based Approach

This paper proposes a method based on Computational Fluid Dynamics (CFD) and the detection of Wind Energy Extraction Latency for a given wind turbine (WT) designed for ultra-short-term (UST) wind energy forecasting over complex terrain. The core of the suggested modeling approach is the Wind Spatial Extrapolation model (WiSpEx). Measured vertical wind profile data are used as the inlet for stationary CFD simulations to reconstruct the wind flow over a wind farm (WF). This wind field reconstruction helps operators obtain the wind speed and available wind energy at the hub height of the installed WTs, enabling the estimation of their energy production. WT power output is calculated by accounting for the average time it takes for the turbine to adjust its power output in response to changes in wind speed. The proposed method is evaluated with data from two WTs (E40-500, NM 750/48). The wind speed dataset used for this study contains ramp events and wind speeds that range in magnitude from 3 m/s to 18 m/s. The results show that the proposed method can achieve a Symmetric Mean Absolute Percentage Error (SMAPE) of 8.44% for E40-500 and 9.26% for NM 750/48, even with significant simplifications, while the SMAPE of the persistence model is above 15.03% for E40-500 and 16.12% for NM 750/48. Each forecast requires less than two minutes of computational time on a low-cost commercial platform. This performance is comparable to state-of-the-art methods and significantly faster than time-dependent simulations. Such simulations necessitate excessive computational resources, making them impractical for online forecasting.

Evaluation of wind energy-based electricity potential at 40m height in 3 districts of Bihar

In this study, three northern districts of Bihar, namely, Bettiah, Madhubani, and Muzaffarpur, were carefully chosen to understand the wind energy resource availability for energy conversion. The wind speed data set was gathered from local meteorological station for a period of 12 years from April 2008 to March 2020. For evaluating wind resources, the widely used two-parameter Weibull distribution was combined with other statistical techniques. The obtained results exhibited that the wind speed in the three districts of Bihar fluctuated from 3.2 m/s to 4.7 m/s while the scale and shape parameters (also known as the Weibull parameters) are discovered to be in the ranges of 10.247 to 11.432 and 3.5718 to 4.0477, respectively. Further, wind power density was 22.729, 26.288, and 28.021 W/m$^2$ at 40 m above the ground for the three districts considered in the study. These findings suggested that the places of the study are not suitable for widespread wind power extraction at a hub height of 40 m but suitable for small-scale wind turbine setup.

Confidence Intervals for the Mean and Difference of Means of Birnbaum-Saunders Distributions with Application to Wind Speed Data

This paper proposes the confidence intervals for the mean and difference of means of Birnbaum-Saunders (BirSau) distributions based on the Bootstrap confidence interval (BCI), Percentile bootstrap confidence interval (PBCI), Generalized confidence interval (GCI), Bayesian credible interval (BayCrI) and the highest posterior density (HPD). The simulation study used R statistical software to evaluate the coverage probabilities and average lengths. The concerning results of the mean suggest that HPD is the recommended method for constructing confidence intervals in the BirSau distributions, except for small sample sizes where the GCI method proves more efficient. For the difference of means, PBCI emerges as the preferred way to construct confidence intervals, except in some cases where small sample sizes with the HPD method are more efficient. Moreover, the average lengths of these proposed confidence intervals decreased as both sample size and shape parameters increased. To illustrate the effectiveness of the suggested confidence intervals, we applied them to wind speed datasets collected in Ayutthaya and Ratchaburi provinces, Thailand.

A wind speed forecasting framework for multiple turbines based on adaptive gate mechanism enhanced multi-graph attention networks

Accurately forecasting wind speed is crucial for efficiently utilizing wind energy and scheduling power grids. Recently, Graph Neural Network (GNN) models have been widely utilized to forecast wind speed, which explicitly utilizes the correlation between turbine sites in a wind farm. However, it is challenging to appropriately construct graphs to characterize multiple latent but unknown interdependencies among turbines. This paper proposes a novel multi-site wind speed forecasting framework AG-MGAT, based on an adaptive gate mechanism enhanced multi-graph attention networks. In detail, the contributions of our work are threefold. Firstly, multiple graphs are explicitly constructed, which respectively measure the wind behavioral similarity and the directional causality between turbine sites. Secondly, to calibrate the potential misalignment of spatial-temporal GNNs using these task-agnostic graphs, an adaptive gate mechanism enhanced Graph Attention Network (GAT), AG-GAT, is innovatively designed, which uses a learnable adjacency matrix as gate to adaptively weight the sites' embeddings from the current GAT layer and these directly from the previous AG-GAT layer. At each timestep, the proposed AG-GATs working on the constructed graphs are used to extract the turbine sites' representations that embed the multiple correlations among sites, which are then sent to recurrent neural network for further processing the temporal interdependency. Finally, thorough experiments on real wind speed dataset are conducted and the experimental results show the superiority of our schemes over other state-of-the-art GNN-based forecasting schemes.

Self-supervised dynamic stochastic graph network for spatio-temporal wind speed forecasting

Spatio-temporal Forecasting has been implemented in diverse fields, such as energy, traffic, and weather. As one typical paradigm of intelligent power systems, spatio-temporal wind speed forecasting has attracted increasing attention. Recently, some graph-based methods have revealed the great potential in the spatio-temporal wind speed forecasting task. Those methods are usually built on a static graph which is incapable of discovering the dynamic characteristics and the uncertainty of temporal wind patterns. To this end, we propose a novel Dynamic Stochastic Graph Network (DSGN) for spatio-temporal wind speed forecasting. Specifically, we characterize each wind turbine of regional power grids with a dynamic stochastic distribution to incorporate the uncertainty into networks. Then Wasserstein distance is adopted to build the stochastic graph, which can effectively represent the spatial dependencies between wind turbines. Additionally, we devise a reconstruction task to preserve the geographical distribution information and regularize the learning of the stochastic graph. To further exploit global spatial dependencies and enhance graph modeling, we augment the stochastic graph view by diffusion operations and integrate a self-supervised learning framework, serving as an auxiliary task, to maximize mutual information between the wind latent states from different graph views. Empirical experiments on four real-world wind speed datasets demonstrate the superiority of DSGN over some state-of-the-art methods. Some ablation studies further validate the effectiveness of the stochastic graph modeling and the self-supervised framework.

Properties and Potentials of Gumbel Power Function Distribution to Rainfall and Wind Speed Datasets

Properties and Potentials of Gumbel Power Function Distribution to Rainfall and Wind Speed Datasets

The Imperial College Storm Model (IRIS) Dataset

Assessing tropical cyclone risk on a global scale given the infrequency of landfalling tropical cyclones (TC) and the short period of reliable observations remains a challenge. Synthetic tropical cyclone datasets can help overcome these problems. Here we present a new global dataset created by IRIS, the ImpeRIal college Storm model. IRIS is novel because, unlike other synthetic TC models, it only simulates the decay from the point of lifetime maximum intensity. This minimises the bias in the dataset. It takes input from 42 years of observed tropical cyclones and creates a 10,000 year synthetic dataset of wind speed which is then validated against the observations. IRIS captures important statistical characteristics of the observed data. The return periods of the landfall maximum wind speed are realistic globally.

A wind speed forecasting model using nonlinear auto-regressive model optimized by the hybrid chaos-cloud salp swarm algorithm

Wind energy forecasting is significantly affected by the strong volatility, intermittency and variability of the wind speed sequence itself. Therefore, to improve forecast accuracy and stability, based on nonlinear auto-regressive model with exogenous inputs (NARX) and the hybrid chaos-cloud salp swarm algorithm (CC-SSA), a short-term wind speed prediction method is proposed. Firstly, to reduce the complexity of the original wind speed data and generate subcomponents with different patterns and low complexity, a mixed modal decomposition method is carried out by combining variational modal decomposition (VMD) based on Pearson correlation coefficient and generalized S-transform (GST) based on adaptive sample entropy. Therefore, the complementary advantages of different mode decompositions are obtained by combining the two different subcomponents into a mixed component. Secondly, by using cloud model and chaotic map, the improved CC-SSA algorithm is proposed to improve the convergence performance of salp swarm algorithm (SSA). Finally, by using CC-SSA algorithm to optimize the weights of NARX, the CC-SSA-NARX forecasting model is established to predict wind speed. The experimental results show that for the 1-Step, 2-Step, and 3-Step datasets of the actual wind speed in the studied region, the MAE, MAPE, RMSE, and R2 of the CC-SSA-NARX model are 0.23, 6 %, 0.27, and 0.97 respectively. The proposed model present the highest prediction index among the other 7 comparative models, showing high accuracy and generalization ability in short-term wind speed prediction, it can provide a certain reference for enhancing the stability of wind power generation and promoting the sustainable development of the wind power industry to some extent.

Evaluation and Improvement of the Accuracy of Reanalysis and Analysis Datasets for Wind Resource Assessment in Sudan

Wind speed datasets are used to evaluate wind resources and energy production of wind farms. In locations where measured data are not available, reanalysis and analysis datasets can be used as an alternative to assess wind resources. This study evaluated the accuracy of wind speed data collected from reanalysis and analysis datasets against mast-measured data between 1975 and 1985 in Sudan, using monthly statistical analyses. Three bias correction methods, based on Measure-Correlate-Predict (MCP) and Linear Adaptation (LA1 and LA2), were applied to determine the original wind speed. The results indicate that LA1 outperformed MCP and LA2. Furthermore, the Weibull distribution function was employed to analyze the wind speed characteristics. In addition, wind power density was calculated using data from different sources. The findings show that although the wind power potential of the chosen locations is not suitable for large wind turbines, wind power can still be exploited with small wind turbines. Consequently, this study introduces a wind energy roadmap to attract investors in clean energy for sustainable development in Sudan, address energy problems, and meet domestic demands. The study also identifies the most important grid datasets for assessing the country's wind potential, enhancing the accuracy of assessments for investors and policymakers.

Facilitating interaction between partial differential equation-based dynamics and unknown dynamics for regional wind speed prediction

Regional wind speed prediction is an important spatiotemporal prediction problem which is crucial for optimizing wind power utilization. Nevertheless, the complex dynamics of wind speed pose a formidable challenge to prediction tasks. The evolving dynamics of wind could be governed by underlying physical principles that can be described by partial differential equations (PDE). This study proposes a novel approach called PDE-assisted network (PaNet) for regional wind speed prediction. In PaNet, a new architecture is devised, incorporating both PDE-based dynamics (PDE dynamics) and unknown dynamics. Specifically, this architecture establishes interactions between the two dynamics, regulated by an inter-dynamics communication unit that controls interactions through attention gates. Additionally, recognizing the significance of the initial state for PDE dynamics, an adaptive frequency-gated unit is introduced to generate a suitable initial state for the PDE dynamics by selecting essential frequency components. To evaluate the predictive performance of PaNet, this study conducts comprehensive experiments on two real-world wind speed datasets. The experimental results indicated that the proposed method is superior to other baseline methods.

Explainable machine learning techniques based on attention gate recurrent unit and local interpretable model‐agnostic explanations for multivariate wind speed forecasting

AbstractWind power has emerged as a successful component within power systems. The ability to reliably and accurately forecast wind speed is of great importance in maintaining the security and stability of the power grid. However, the significance of explaining prediction models has often been overlooked by researchers. To address this gap, this study introduces a novel approach to wind speed forecasting that incorporates a significant decomposition method, attention‐based machine learning, and local explanation techniques. The proposed model utilizes grid search variational mode decomposition to decompose the wind speed sequence into different modes while employing gate recurrent unit with an attention mechanism to achieve superior forecasting performance. Experimental evaluations conducted on eight real‐world wind speed datasets demonstrate that the proposed approach outperforms other popular models across multiple performance criteria. In two specific experiments, the proposed approach achieved a minimal mean absolute percentage error of 2.74% and 1.70%, respectively. Furthermore, local interpretable model‐agnostic explanations (LIME) were employed to assess the influence of factors, highlighting whether they positively or negatively affected the predicted values.

A wind speed forcasting model based on rime optimization based VMD and multi-headed self-attention-LSTM

Due to the nonlinearity, fluctuation, and intermittency of wind speed, its accurate prediction is essential for improving efficiency in wind power operation systems. In this regard, a hybrid model that combines the rime optimization algorithm (RIME), variational mode decomposition (VMD), multi-headed self-attention (MSA) mechanism and long short-term memory (LSTM) is proposed for wind speed prediction. First, the number of modes and VMD penalty parameter are optimized with RIME, the optimized parameters are brought into the VMD to decompose the raw wind speeds, and a Lagrange multiplier and quadratic penalty function are introduced to obtain the input series. Then, a LSTM short-term wind speed prediction model is constructed based on the MSA mechanism and solved for the hidden states and weights of each layer of attention in the model. Finally, a ReLU activation function is used to activate the hidden states of the LSTM model, and a weighted sum vector is used as the final sequence representation, which is inputted to the output layer for specific prediction to obtain the short-term wind speed prediction results. To verify the effectiveness of the proposed model, wind speed data from four wind farms in Ningxia, China, and two sets of wind speed data from an M2 tower in the USA are selected, and 19 models are built to compare the performance of the proposed model. The results show that the proposed model outperforms other models on all datasets in terms of all five performance metrics, with smaller errors and higher prediction accuracy.

Wind speed forecasting using Singular Systems Analysis

We report on a method to predict wind speeds up to 24 hours ahead using a technique originating in Dynamical Systems and Chaos theory using a signal processing technique known as Singular Systems Analysis. The method predicts wind speeds based on a set of previous measurements which were used to construct an attractor in an optimally defined phase space as a ‘training set’. Current wind measurements can then be projected to onto that phase space to find most similar previous measurements. By tracing the evolution of these similar previous data, it is possible not only to forecast the wind speed but also to obtain a measure of the expected forecasting uncertainty. The method was applied to a set of hourly wind speed data from a UK Meteorological Office weather station near Edinburgh. A comparison with a simple persistence prediction showed that the Singular Systems Analysis was both, consistently better at predicting wind speeds between 12 and 24 hours ahead than persistence, and also able to provide a meaningful forecasting uncertainty.

Evaluation of ERA5 and MERRA-2 Reanalysis Datasets over the Aegean Region, Türkiye

Reanalysis products are among the most-used datasets in the atmospheric sciences since they comprehensively describe the observed climate at sub-daily intervals in a region. Two reanalysis datasets, namely, the fifth generation of European Centre for Medium-range Weather Forecast (ECMWF) atmospheric reanalysis of global climate (ERA5) and Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA2), were evaluated for the representation of air temperature at 2 m, mean sea level pressure and wind speed over the Aegean Region of Türkiye during the period 1963–2020. Hourly reanalysis data were compared with observations in 19 meteorological stations in the region. Several statistical parameters, such as root mean square error (RMSE), correlation coefficient (R), and mean bias error (MBE), were used to evaluate the performances of the datasets. The results indicated that air temperature and mean sea level pressure are generally better represented by the MERRA-2 reanalysis in the region, whereas the ERA5 reanalysis dataset better represents wind speed. MERRA-2 had lower RMSE and slightly better performance at 11 stations with high R (>0.98) for mean sea level pressure. The MERRA-2 reanalysis dataset had a high overall R (>0.94) for air temperature and performed better at 12 stations. The overall regional R-value for the ERA5 wind speed dataset was 0.58, and ERA5 showed better performance at 13 individual stations for wind speed. Our results guide which reanalysis dataset better represents the regional climate characteristics for selected parameters.

An innovative interpretable combined learning model for wind speed forecasting

Wind energy is taken as one of the most potential green energy sources, whose accurate and stable prediction is important to improve the efficiency of wind turbines as well as to guarantee the power balance and economic dispatch of power systems and equipment safety. However, the random and fluctuating nature of wind speed poses a great risk to wind power grid connections. To address the issues of low prediction performance and lack of interpretable analysis in most past studies, this research proposes an interpretable combined learning model for wind speed time series prediction by combining linear models, different neural networks, and deep learning by introducing interpretable TFT models. To test the effectiveness of the forecasting models, the presented combined model is verified using eight wind speed datasets covering four seasons collected from two wind farms in Shaanxi, China. The experimental results show that the average root mean squared error of the one-step, two-step, and three-step predictions on the eight datasets for proposed model are 0.3448, 0.4586 and 0.6164, respectively, which are much better than the six single models and the six combined models with different strategies. And proposed model outperforms the single model and combined model in most cases, with 86.80% and 92.01% of the DM values greater than the corresponding critical values when the significance level is set to 0.01 and 0.1, respectively. Finally, the proposed model is discussed and analyzed in depth through interpretability analysis of the combined model, which further validates the potential of the model and also provides a reference for other time series forecasting studies.

Risk assessment and spatio-temporal characteristics analysis of water bloom in three large-scale eutrophic reservoirs in Fujian Province, China

The occurrence of water blooms in reservoirs poses a serious threat to water safety, so the assessment of water bloom risk is crucial. In this study, we constructed a reservoir water bloom risk index (RWBRI) using the trophic state index (TSI), floating algae index (FAI), temperature, precipitation, and wind speed datasets, and estimated and validated the water bloom risk levels of three large-scale eutrophication reservoirs in Fujian Province, China, from 2019 to 2023 based on the Sentinel-2 MSI dataset at the pixel scale. The results showed that there is no positive correlation between the TSI and FAI parameters of the three reservoirs (the significance test coefficients for Shanzai Reservoir, Dongzhen Reservoir, and Shanmei Reservoir were −0.455**, −0.371**, and −0.049, respectively, and ** indicates that the reservoir passed the statistical significance test at the 0.01 significance level). The TSI of Shanmei Reservoir was less fluctuating and more stable than that of Shanzai Reservoir and Dongzhen Reservoir, but the average value of FAI was higher. The average value of water bloom risk in Shanzai Reservoir and Dongzhen Reservoirs is higher than Shanmei Reservoir (0.47, 0.46, and 0.41, respectively), but the trend of annual average changes indicates that the overall risk of Dongzhen Reservoir is decreasing, while the risk of Shanzai Reservoir and Shanmei Reservoir is increasing. Due to the influence of weather, the water bloom risk in spring and summer is higher than that in fall and winter in the three reservoirs, and the highest risk value usually occurs in summer. The high values in Shanzai Reservoir were mainly distributed in the two inlet channels in the northern and southwestern parts of the reservoir, the bank risk was higher than in the other areas of Dongzhen Reservoir, and the high value of the risk in Shanmei Reservoir was mainly distributed in the center of the reservoir. The RWBRI method can be used for water bloom risk identification in reservoirs with eutrophication trends by comparing it with four-hourly and monthly water quality level data.

Enhancing the accuracy of wind speed estimation model using an efficient hybrid deep learning algorithm

Estimation of wind speed holds great significance for efficient operation of wind farms and grid stability. Randomness of wind speed poses a great challenge for accurate estimation of wind speed. Hence for addressing this problem, a convolutional neural network (CNN) and long short-term memory (LSTM) based novel hybrid deep learning (DL) short-term is proposed. The prediction model is built using pre-processed wind speed dataset from Jaisalmer district, Rajasthan. For designing the proposed model, the samples of the dataset are bifurcated into four seasons such as autumn, summer, winter and spring and the prediction performance are evaluated for each season separately. The proposed hybrid algorithm (CNN-LSTM) is implemented and a comparison is made with different DL algorithms. The predictive performance are calculated by observing the statistical metrics such as root mean squared error (RMSE), mean absolute error (MAE), relative RMSE, normalized RMSE and coefficient of determination (R2). It is observed that the value of statistical metrics such as MAE, RMSE, relative RMSE and normalized RMSE of the proposed algorithm are lower than the other DL models for all the four seasons. The higher value of R2 obtained by the CNN-LSTM indicates the model is best fitted with the dataset. Conclusively, the proposed model demonstrates improved adaptability and higher prediction accuracy in comparison to the predictive skills of models utilizing the same dataset.