Wind Speed Datasets Research Articles

A short-term hybrid wind speed prediction model based on decomposition and improved optimization algorithm

Introduction: In the field of wind power generation, short-term wind speed prediction plays an increasingly important role as the foundation for effective utilization of wind energy. However, accurately predicting wind speed is highly challenging due to its complexity and randomness in practical applications. Currently, single algorithms exhibit poor accuracy in short-term wind speed prediction, leading to the widespread adoption of hybrid wind speed prediction models based on deep learning techniques. To comprehensively enhance the predictive performance of short-term wind speed models, this study proposes a hybrid model, VMDAttention LSTM-ASSA, which consists of three stages: decomposition of the original wind speed sequence, prediction of each mode component, and weight optimization.Methods: To comprehensively enhance the predictive performance of short-term wind speed models, this study proposes a hybrid model, VMDAttention LSTM-ASSA, which consists of three stages: decomposition of the original wind speed sequence, prediction of each mode component, and weight optimization. Firstly, the model incorporates an attention mechanism into the LSTM model to extract important temporal slices from each mode component, effectively improving the slice prediction accuracy. Secondly, two different search operators are introduced to enhance the original Salp Swarm Algorithm, addressing the issue of getting trapped in local optima and achieving globally optimal short-term wind speed predictions.Result: Through comparative experiments using multiple-site short-term wind speed datasets, this study demonstrates that the proposed VMD-AtLSTM-ASSA model outperforms other hybrid prediction models (VMD-RNN, VMD-BPNN, VMD-GRU, VMD-LSTM) with a maximum reduction of 80.33% in MAPE values. The experimental results validate the high accuracy and stability of the VMD-AtLSTM-ASSA model.Discussion: Short-term wind speed prediction is of paramount importance for the effective utilization of wind power generation, and our research provides strong support for enhancing the efficiency and reliability of wind power generation systems. Future research directions may include further improvements in model performance and extension into other meteorological and environmental application domains.

Short-term wind speed forecasting based on adaptive secondary decomposition and robust temporal convolutional network

Accurate and reliable short-term wind speed forecasting remains challenging due to the high volatility and randomness of wind speed. The secondary decomposition (SD) method deeply extracts the entangled fluctuation patterns by leveraging the advantages of two distinct decomposition methods. Nevertheless, two critical issues remain to be further explored, including (1) how to adaptively select the sub-signals to be further decomposed and (2) how to effectively capture the coupling temporal dependencies among the sub-signals. To address these issues, we propose a novel hybrid model based on an adaptive secondary decomposition (ASD) method and a robust temporal convolutional network (RTCN). Firstly, the proposed ASD method is developed to adaptively extract more sub-signals with lower complexities by combining the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), entropy-based spectral clustering (ESC), and the variational mode decomposition (VMD). Subsequently, a single RTCN is fitted to capture the temporal dependencies among the sub-signals and identify the global relationship between the sub-signals and the future wind speed. The forecasting performance is verified on four real-world wind speed datasets from different wind farms, and the experimental results demonstrate that the proposed ASD-RTCN model consistently outperforms the benchmark models in terms of forecasting accuracy and stability.

Reconstruction of Near‐Surface Hourly Wind Speed Data Sets During the Period 1959–2021 Over China

AbstractThe reconstruction models of hourly 10‐m wind speeds were developed for each of 2384 stations over China using stepwise regression, random forest and XGBoost machine learning approaches based on hourly observed and ERA5 reanalysis data from 2005 to 2021. The reconstruction procedures applied observed hourly data to reduce the systematic biases of reanalysis data sets. Furthermore, the procedures employed the past dynamically consistent states of the atmosphere simulated by ERA5 reanalysis techniques to reduce/remove wind‐observed data impacts of long‐term non‐meteorological condition changes over time (e.g., urbanization around weather stations), which provide homogenous hourly wind speed data sets from 1959 to 2021. The systematic errors of the models' simulations derived from three approaches are similarly small, with almost two orders of magnitude smaller than ERA5 original data sets. The systematic errors of reconstructed data sets derived from stepwise regression are similar to its simulation; however, the biases from two machine learning methods are even much greater than ERA5 original data sets. This result implies that machine learning methods are not suitable for such typical time‐series predictions using the previous‐hour wind speed as a predictor to reconstruct wind speed data for the next hour. Therefore, stepwise regression was selected to reconstruct hourly wind speed data sets, which have much better quality than ERA5 reanalysis data with the median correctness increased by >50% and the median rRMSE decreased by 25%–50%. Consequently, the reconstructed wind speed data sets have great potential to be useful for more precisely assessing the characteristics/trends of wind energy resources in the past 60 years over China.

On the development of new cosine-based probabilistic methods with applications to univariate and bivariate analyses of the wind speed energy

So far in the literature, a number of probability distributions have been successfully implemented for analyzing the wind speed and energy data sets. However, there is no published work on modeling and analyzing the wind speed and energy data sets with probability distributions that are introduced using trigonometric functions. In the existing literature, there is also a lack of studies on implementing the bivariate trigonometric-based probability distributions for modeling the wind speed and energy data sets. In this paper, we take up a meaningful effort to cover these interesting research gaps. Thus, we first incorporate a cosine function and introduce a new univariate probability distributional method, namely, a univariate modified cosine-G (UMC-G) family. Using the UMC-G method, a new probability distribution called a univariate modified cosine-Weibull (UMC-Weibull) distribution is studied. We apply the UMC-Weibull distribution for analyzing the wind energy data set taken from the weather station at Sotavento Galicia, Spain. Furthermore, we also introduce a bivariate version of the UMC-G method using the Farlie–Gumble–Morgenstern copula approach. The proposed bivariate distributional method is called a bivariate modified cosine-G (BMC-G) family. A special member of the BMC-G distributions called a bivariate modified cosine-Weibull (BMC-Weibull) distribution is introduced. We apply the BMC-Weibull distribution for analyzing the bivariate data set representing the wind speed and energy taken from the weather station at Sotavento Galicia. Using different statistical tools, we observe that the UMC-Weibull and BMC-Weibull are the best-suited models for analyzing the wind speed and energy data sets.

GAN-Based Abrupt Weather Data Augmentation for Wind Turbine Power Day-Ahead Predictions

This study introduces a data augmentation technique based on generative adversarial networks (GANs) to improve the accuracy of day-ahead wind power predictions. To address the peculiarities of abrupt weather data, we propose a novel method for detecting mutation rates (MR) and local mutation rates (LMR). By analyzing historical data, we curated datasets that met specific mutation rate criteria. These transformed wind speed datasets were used as training instances, and using GAN-based methodologies, we generated a series of augmented training sets. The enriched dataset was then used to train the wind power prediction model, and the resulting prediction results were meticulously evaluated. Our empirical findings clearly demonstrate a significant improvement in the accuracy of day-ahead wind power prediction due to the proposed data augmentation approach. A comparative analysis with traditional methods showed an approximate 5% increase in monthly average prediction accuracy. This highlights the potential of leveraging mutated wind speed data and GAN-based techniques for data augmentation, leading to improved accuracy and reliability in wind power predictions. In conclusion, this paper presents a robust data augmentation method for wind power prediction, contributing to the potential enhancement of day-ahead prediction accuracy. Future research could explore additional mutation rate detection methods and strategies to further enhance GAN models, thereby amplifying the effectiveness of wind power prediction.

A hybrid VMD based contextual feature representation approach for wind speed forecasting

Accurate wind speed prediction is critical for efficient power system operation, regulation, security analysis, and energy trading. However, the stochastic nature of the wind makes wind speed forecasting (WSF) difficult. Thus, a novel hybrid WSF approach termed VMD-Ts2Vec-SVR comprising variational mode decomposition (VMD), contextual time series representation (Ts2Vec) model, and support vector regression (SVR) is proposed. In the proposed approach, VMD is used to decompose the raw input wind speed for denoising, and extracting the main features of the original series, Ts2Vec model is used to learn the sequential contextual representations in all semantic levels from the denoised series, and SVR is used to predict the future wind speed from the contextual representation. Two experiments are performed for testing the proposed approach using wind speed dataset collected from Leicester, and Portland wind farms. For validation of the proposed approach for different time intervals, it is tested for 5-min, 10-min, 15-min, 30-min, 1-h, and 2-h ahead WSF. The performance of the proposed approach is compared with seven individual models, seven hybrid VMD based models for better validation. Two experiments demonstrated both the proposed approach’s superior performance across all time horizons and its viability for the WSF.

Assessing the performance of several numerical methods for estimating Weibull parameters for Wind Energy Applications: A case study of Al-Hodeidah in Yemen

The estimation of wind potential is generally assessed using the two parameters of Weibull distribution (k, c). Relative Percentage Error (RPE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Relative Root Mean Square Error (RRMSE), Correlation Coefficient (R), and Index of Agreement (IA) are calculated to comparatively analyze five numerical methods of computing Weibull parameters. The five selected methods are (Empirical Justus (EMJ), Energy Pattern Factor (EPFM), Maximum Likelihood (MLM), Method of Moment (MOM), and L-moment Estimation Method (L-MOM)). With the daily wind speed dataset for the year 2014 measured at 10 m height, the five methods have been applied on the study site Al-Hodeidah city in Yemen. Therefore, compared to other methods, the EPFM method is the most accurate method for estimating the Weibull parameters, while the L-MOM method is the least accurate at 10, 30, and 50 m heights. Moreover, this study reveals that the parameter remains the same with height, while the c parameter increases with height. The values of the skewness and kurtosis of wind speed and power density do not change as height changes.

The properties of the global offshore wind turbine fleet

Offshore wind capacity increased massively in recent years. Sufficient wind potential must be available to realize the ambitious goals regarding further wind capacity expansion. Thus, the goal of this study is to quantify the meteorological, geographical, and technical properties of the current global offshore wind turbine fleet. The offshore wind turbine fleet is classified by twelve factors that belong to meteorological, geographical, and technical factor categories. A comprehensive wind turbine site dataset is evaluated, including 5473 sites in Europe and 3404 in Asia. The factors at the wind turbine sites are derived using publicly available datasets, including a reanalysis wind speed dataset, geodata, and technical properties. The results of this study are that the global median (1) wind speed is 8.7 m/s, (2) water depth is 17 m and (3) distance to shore is 27 km. The distance to the nearest wind turbine is 5.1 times the rotor diameter. The factors studied have a high regional variability. While Asian offshore wind turbines operate in shallower water closer to the shores than European, European wind turbine sites provide higher wind resources. The regional differences indicate that the wind potential and wind turbine siting criteria differ depending on the country. The factor values obtained are reference values for future offshore wind resource assessments and estimates of the wind potential from a regional to global scale.

Rolling decomposition method in fusion with echo state network for wind speed forecasting

Accurate wind speed forecasting is beneficial to ensure the safe and stable operation of power systems, improve economic benefits, and promote the healthy development of the wind power industry. This study develops a novel hybrid model called VMD-ESN-STO combining the rolling variational mode decomposition (VMD), echo state network (ESN), and subseries to original series (STO) structure for wind speed forecasting. In this model, the rolling method is not only used in decomposition, but also in training and forecasting. The rolling VMD is used to decompose the original wind speed series into several subseries according to the rolling schema, ESN is used to forecast, and STO structure determines the input and output of the forecasting model. Four wind speed datasets are utilized for wind speed forecasting experiments to validate the applicability and accuracy of the developed model. Mean absolute percentage errors of VMD-ESN-STO in the four datasets are 4.4511%, 2.4451%, 4.1400%, and 2.6178%, respectively, which are far less than the errors for the six comparative models. The developed VMD-ESN-STO is an appropriate tool for wind speed forecasting due to its superior forecasting performance.

Evaluating the Impact of Weather Conditions on the Effectiveness and Performance of PV Solar Systems and Inverters

Solar photovoltaic (PV) systems have become an increasingly popular and environmentally friendly source of renewable energy. However, the performance and effectiveness of these systems can be significantly influenced by various weather conditions. This paper aims to evaluate the impact of weather conditions on the effectiveness and performance of PV solar systems and inverters. This paper utilizes a comprehensive dataset of solar irradiance, temperature, wind speed, and other weather parameters collected from multiple locations over an extended period. Various performance indicators, such as power output, efficiency, and reliability, are analyzed and compared under different weather conditions. The findings of this paper reveal the significant influence of weather conditions on the performance of PV solar systems and inverters. It is observed that variations in solar irradiance, temperature can lead to fluctuations in power output and efficiency. High ambient temperatures, for example, can negatively affect the performance of solar panels by increasing the operating temperature and reducing their conversion efficiency. These events can cause physical damage, reduce solar irradiance, and impair the functionality of inverters, leading to a decrease in overall system performance. The research outcomes presented in this paper provide valuable insights for system designers, engineers, policymakers, and investors in the renewable energy sector.

The Role of Teleconnection Patterns on Wind Speed of Iraqi-Kurdistan Region

Changes in the wind speed of the Iraqi-Kurdistan region are associated with many types of teleconnection patterns. Daily wind speed datasets for eight meteorological stations in the Iraqi-Kurdistan region were gathered from Sulaymaniya, Penjwen, Hallabja, Kalar, Erbil, Sallahadin, Zakho, and Duhok weather stations. spectral comparability between the daily wind speed of the eight sites and prominent atmospheric indices was implemented for the North Atlantic Oscillation (NAO), Southern Oscillation (SO), Mediterranean Oscillation (MO), and Arctic Oscillation (AO), using the multi-taper method (MTM) for spectral analysis to find the source of the wind speed cycle of the cites for the study area. It was found that the common cycles for the eight stations are 2.22, 2.5, 3.33, 3.7, 4.35, 5.56, 7.14, 10, and 14.29 cycles. It is obvious that there are common cycles between each of the NAO, SO, MO, and AO indices and the cycles of daily wind speed of the eight stations. The results of this study will allow us to understand the relation between the regional daily wind speeds patterns fluctuations and the global teleconnection patterns and also to forecast the behavior of daily regional wind speed.

Estimating average wind speed in Thailand using confidence intervals for common mean of several Weibull distributions.

The Weibull distribution has been used to analyze data from many fields, including engineering, survival and lifetime analysis, and weather forecasting, particularly wind speed data. It is useful to measure the central tendency of wind speed data in specific locations using statistical parameters for instance the mean to accurately forecast the severity of future catastrophic events. In particular, the common mean of several independent wind speed samples collected from different locations is a useful statistic. To explore wind speed data from several areas in Surat Thani province, a large province in southern Thailand, we constructed estimates of the confidence interval for the common mean of several Weibull distributions using the Bayesian equitailed confidence interval and the highest posterior density interval using the gamma prior. Their performances are compared with those of the generalized confidence interval and the adjusted method of variance estimates recovery based on their coverage probabilities and expected lengths. The results demonstrate that when the common mean is small and the sample size is large, the Bayesian highest posterior density interval performed the best since its coverage probabilities were higher than the nominal confidence level and it provided the shortest expected lengths. Moreover, the generalized confidence interval performed well in some scenarios whereas adjusted method of variance estimates recovery did not. The approaches were used to estimate the common mean of real wind speed datasets from several areas in Surat Thani province, Thailand, fitted to Weibull distributions. These results support the simulation results in that the Bayesian methods performed the best. Hence, the Bayesian highest posterior density interval is the most appropriate method for establishing the confidence interval for the common mean of several Weibull distributions.

Short-term wind speed prediction based on fractal dimension-variational mode decomposition and general continued fraction

Accurate wind speed prediction is significant to maintain the stable power system operation and enhance wind power utilization efficiency. However, most of wind speed prediction models cannot well fit the variation pattern of wind series due to its random and chaotic characteristics. This study proposes a novel wind speed prediction model incorporating fractal dimension (FD) and variational mode decomposition (VMD) and general continued fraction (GCF). More specifically, the fractal feature of wind speed series is analyzed, and the parameter determination strategy for VMD based on fractal feature of wind series is developed to overcome the problem of ambiguous mode parameter K for VMD, so that the relatively regular modes can be extracted from the raw wind series. To well capture the variation characteristics of wind series, a novel GCF model is derived on the basis of the inverse difference quotient theory, and the structure parameters of GCF model are determined by the bats algorithm (BA). To verify the accuracy and stability of the proposed model, several experiments involving comparisons with some mainstream models are executed on five wind speed datasets, and the proposed model outperforms comparative models with mean absolute percentage error reduction of 0.48 %–74.44 %. The results of the experiments indicate the proposed model has great capacity of capturing the variation characteristics of wind speed series.

A Novel Method of Forecasting Chaotic and Random Wind Speed Regimes Based on Machine Learning with the Evolution and Prediction of Volterra Kernels

This study aims to focus on using the Volterra series and machine learning for forecasting random and chaotic wind speed regimes, since calm weather is mostly noticed at the local site, making dataset selection difficult. A novel method is proposed to predict Volterra kernels up to the third order, using a forward–back propagation neural network with 12-month measurements at Fujairah site (United Arab Emirates). Both daily and monthly wind speed datasets are investigated for forecasting. The three dominant hourly and daily kernels are extracted for each day and each month. Predicted future Volterra kernels are estimated from past values using both statistical analysis and individual neuro networks for each of the Volterra kernel coefficients. Using the evolved Volterra kernels, the hourly and daily wind speeds are forecasted with similar patterns of the measured values. Due to the random nature of wind speed at the local site, a two-layer with four neurons per layer neuro network is used to locate the most variable and intense speed during 8 h in the day. Forecasted wind speed is determined with errors arising from different sources, such as the utilization of only third-order Volterra kernels and the difficulty of machine training of the employed shallow network. Nevertheless, this work depicts a useful algorithm to forecast chaotic and random wind speed regimes. Computational time is a trade of the complexity of Volterra mathematical analysis.

Multi-step wind speed prediction based on LSSVM combined with ESMD and fractional-order beetle swarm optimization

Accurate and stable wind speed prediction can alleviate the uncertain impacts of wind power generation caused by nonlinear characteristics of wind speed, and then improve the reliability of wind power. In this paper, a hybrid model for wind speed prediction based on mode decomposition, parameter optimization and basic prediction model is proposed. First, the extreme-point symmetric mode decomposition (ESMD) is employed to adaptively decompose the denoised wind speed time series into sub-sequences with different frequencies. Second, a fractional-order beetle swarm optimization (FO-BSO) for parameter optimization of the Least squares support vector machine (LSSVM) is proposed. Through benchmark functions and non-parametric statistical test, the advantages of the FO-BSO in accuracy, stability and convergence speed are verified. Subsequently, the ESMD-FO-BSO-LSSVM prediction model is established, and three groups of wind speed datasets with different sampling locations and sampling frequencies are selected for simulation experiments. The results show that the coefficient of determination of 1-step prediction of the proposed model in three datasets are 0.9856, 0.9713, 0.9940, which has 2.43%, 3.38%, 3.08% average promotion than that of 7 comparative models. And the accuracy and stability of ESMD-FO-BSO-LSSVM model in multi-step wind speed prediction have also achieved better performance than 7 competitors.

A general unit hydrograph distribution and its application on the marginal distribution of global wind speed

We here introduce a simple, general, and analytical method based on the general unit hydrograph (GUH) theory proposed by Guo (2022a, 2022b, 2022c) to describe wind speed marginal distribution. Wind speed probability is generally described by the Weibull distribution; however, its accuracy depends on the contribution made by null wind speed (owing to its zero initial condition). To overcome this disadvantage, we propose a three-parameter GUH distribution with significant flexibility in modeling symmetric heavy-tailed distributions as well as skewed distributions. We show that our proposed GUH distribution corresponds to a generalized logistic distribution that contains both type I and type II generalized logistic distributions. The shape properties of the GUH distribution are discussed along with a simple method for estimating parameters by means of least squares methods. The flexibility of the proposed GUH distribution was assessed by applying it to a well reconstructed global gridded wind speed dataset. Subsequently, we quantify the changes caused by a reversal in global surface wind speed stilling using our GUH distribution. Finally, we confirm that our proposed GUH model agrees with reconstructed wind speed data better than conventional Weibull distribution models, which suggests that the new distribution can be a useful tool for sustainable development of wind energy.

Downscaling wind speed based on coupled environmental factors and machine learning

AbstractWind speed changes impact society and have important implications for climate change studies. Thus, high‐resolution and high‐quality wind speed datasets are necessary for environmental monitoring and ecosystem research. However, there is no complete set of high spatial and temporal resolution wind speed datasets for China. Additionally, it is extremely challenging to produce wind speed data at high spatial and temporal resolution for large‐scale regions with diverse climate types and complex topographies, such as China. In this study, we used multisource remote sensing images, obtained data on various environmental factors through the Google Earth Engine and Evapotranspiration (ET) Watch Cloud platforms, and combined machine learning algorithms to downscale the ERA5 reanalysis wind speed data, and finally obtained the daily wind speed datasets with 1 km spatial resolution for China in 2015. To verify the accuracy of the model and data products, we selected several metrics to evaluate in conjunction with the actual site observed data. The results show that the multifactor combination model of artificial neural network combining land surface temperature, sunshine durations and roughness factors outperforms a single‐factor combination model, and the results were in good agreement with the original data (R2 of 0.95 and RMSE of 0.40 m·s−1). The final wind speed data products were also in good agreement with the observed meteorological data (R2 range of 0.86–0.95 and RMSE range of 0.33–0.44 m·s−1); moreover, the accuracy and precision were greatly improved over the original data. This study provided a dataset that has potential applications in future climate change and ecosystem studies.

Comparison of the performance of different wind speed distribution models applied to onshore and offshore wind speed data in the Northeast Brazil

This paper aims to evaluate the performance presented by different distribution models when applied to onshore and offshore wind speed data and determine which distribution has the best fit in each location. The onshore data were measured at two stations located in the Northeast Brazil and the offshore data were measured by two ocean buoys located in the South Atlantic. Five probability distributions were used to model wind speed, namely: Weibull, Nakagami, extended generalized Lindley, generalized gamma and generalized extreme value. The probability distribution parameters were estimated using maximum likelihood, modified maximum likelihood and the method of multi-objective moments. In addition, three goodness of fit tests were used to select the distribution model that best fits a region’s wind speed data, namely: Kolmogorov–Smirnov test, Deviation of Skewness and Kurtosis and the Akaike Information Criterion. The results of these tests were aggregated into a single performance metric called “total error”. In both onshore locations, the extended generalized Lindley and generalized gamma were, in general, superior to the other distributions. In the offshore wind data from the ocean buoys, the extended generalized Lindley, generalized gamma and Weibull showed the best fit in adjusting the histogram. The generalized extreme value distribution presented the worst fit for all four locations. According to the results, there is no single distribution model that is more appropriate to fit a set of wind speed data, and it is necessary to carry out a study to know which of the available distributions is the most appropriate for any particular dataset.

An imputation and decomposition algorithms based integrated approach with bidirectional LSTM neural network for wind speed prediction

Due to its renewable and ecological attributes, wind energy is receiving attention on a global scale. However, exact forecasting of wind speed is challenging owing to its variable and stochastic nature. Further, many algorithms established for wind speed forecasting has neglected the missing value imputation study. In fact, the prediction accuracy of any forecasting system may suffer significantly if any missing data occurs during forecasting due to power off of any unexpected device, failure of communication/transmission equipment or sensor, measurement error, or other unknown causes. In order to take care of this, first k-NN imputation algorithm is used. Thereafter, in the proposed framework, the characteristics of several data decomposition approaches namely EMD, EEMD & CEEMDAN are integrated with BiDLSTM Neural Network for de-noising the signal. For testing the accuracy of the proposed forecasting technique, this study investigates several wind speed datasets gathered from NIWE data portal for 5 cities i.e. Chandori, Bhogat, Gandhinagar, Charanka and Surat situated in the Gujarat state (India). The empirical finding demonstrates that the proposed k–NN–CEEMDAN-BiDLSTM based hybrid technique in terms of prediction accuracy outperforms other hybrid techniques exist in literature.

Dynamic adaptive spatio-temporal graph neural network for multi-node offshore wind speed forecasting

Multi-node offshore wind speed forecasting is a challenging task due to the complex dynamic spatial dependencies and highly nonlinear temporal dynamics present in the ocean. As deep learning advances, graph neural networks (GNNs) have great potential to capture spatial dependencies in ocean meteorology. However, existing GNN models usually use predefined or learned static graphs. They lack the ability to model dynamic spatial associations, which can limit the performance of GNNs. In this paper, we propose a dynamic adaptive spatio-temporal graph neural network (DASTGN) that uses dynamic graph convolution (DGCN) to capture dynamic spatial dependencies in offshore wind speed data. Based on the assumption that not only long-term static associations but also short-term dynamic associations exist in the spatial domain and that the importance of these two associations is different, we propose a dynamic adaptive graph generation module to generate static and dynamic graphs to model these two associations. Meanwhile, a matrix fusion mechanism is proposed to fuse them into the optimal dynamic graph, which is fed into the DGCN module. We employ a temporal convolution module to capture the nonlinear temporal dependencies. Finally, the above modules are integrated into a dedicated spatio-temporal convolution module to predict wind speed. Extensive experiments on real wind speed datasets in Chinese seas showed that the DASTGN improved the performance of the optimal baseline model by 3.05% and 3.69% in terms of the MAE and RMSE, respectively. To demonstrate that the DASTGN can effectively model dynamic spatial associations, the generated graph structure is visualized and analyzed. Finally, we present policy implications aimed at enhancing the security of the power system.