Wind Speed Series Research Articles

Spatio-temporal wind speed prediction based on Clayton Copula function with deep learning fusion

Accurate prediction of wind speed plays an important role in increasing the power generation of wind turbines and realizing efficient use of wind energy. However, due to the large number of wind turbines in the wind farm and the complex wake effects between the units, the coupling degree and spatial correlation of the wind speed of the wind turbines are increased. Accordingly, this paper proposes a wind speed prediction model based on spatio-temporal dependency analysis. The proposed model first uses long short-term memory(LSTM) neural network to predict the wind speed of each wind turbine to obtain its residuals, which can extract the time correlation of the wind-speed series; Then by applying the Clayton Copula function to analyze the correlation between the residual series and wind-speed series to get the joint-distribution function. The joint-distribution function can be used to calculate the prediction error of the wind speed and complete the wind speed prediction. The validity of the method in this work is verified using the measured wind-speed data of a wind farm. Experimental results show that the method effectively solves the problem of finding spatio-temporal dependency of wind speed and significantly improves the prediction accuracy of wind speed.

Robustified extreme learning machine regression with applications in outlier-blended wind-speed forecasting

Wind energy is a core sustainable source of electric power, and accurate wind-speed forecasting is pivotal to enhancing the power stability, efficiency, and utilization. The existing forecasting methods are still limited by the influence of outliers and the modelling difficulties caused by complex features in wind speed series. This paper proposes a new wind speed forecasting system based on a designed adaptive robust extreme learning machine (ARELM) model and signal decomposition algorithms. Firstly, the ARELM is designed to sufficiently lessen the violation of normality assumptions and contamination by outliers. ARELM takes an adaptive scaled Huber’s loss as its objective function, which can limit the influence of outliers and adaptively determine an appropriate mixture distribution of normal distribution and Laplace distribution at the same time. Secondly, the empirical mode decomposition (EMD) method and its improved methods (EEMD, CEEMD and CEEMDAN) are introduced to our wind-speed forecasting system, where the low-frequent sub-series are modelled by basic ELM and the high-frequent ones are modelled by ARELM. This can decompose the modelling complex wind speed series into modelling several simple sub-series and reduce the difficulty of modelling. Experimental results show that our combined forecasting system, ELM-ARELM, obtains up to 78% improvement in forecasting performance comparing with the methods using general Huber’s loss and other comparison methods, which show the superiority of the adaptive scaled Huber’s loss. The error indexes (MAE and RMSE) by the proposed system, which are (0.25, 0.34), (0.32, 0.45) and (0.38, 0.53) for 5 min head, 15 min ahead and 25 min ahead experiments respectively, demonstrate the effectiveness of decomposition methods on improving accuracy of wind speed prediction.

The annual rate of independent events – A key interpretation for traditional extreme value distributions of wind velocity

AbstractThe extreme value theory has been object of engineering studies for more than a century. The analysis of extreme winds plays a key role for complex civil structures and a driving role in different stages of wind turbines lifetime. Most of extremes probability models depend on the annual rate of independent events (ARIE) which has been traditionally considered a constant value. The authors have embraced a recent belief considering the ARIE as a function of the wind velocity. Even though a certain agreement has been achieved across the researches, some issues are still pending. In this regard, the paper shows that the annual, seasonal and daily fluctuations embedded in time series of the mean wind speeds, constrain its probability distribution and time correlation to be physically consistent. Besides, a new physical interpretation of the ARIE is presented, expressing how the independence across wind observations increases with the wind speed, up to the point that all yearly observations are independent if larger than a suitable speed value. Such a tendency is not revealed if the annual, seasonal and daily fluctuations are excluded by the analysis, leading to a deceitful shape of the ARIE. Finally, the paper shows how the velocity‐dependent ARIE model is consistent with the conventional asymptotic extreme value theory, if a sufficiently large left‐censorship applies to the dataset. The study of the ARIE presented in this paper is based on long‐term Monte Carlo simulation of the mean wind speed.

Wind stilling ceased in the Iberian Peninsula since the 2000s

This study analyses quality controlled and homogenized near-surface wind speed series (SWS) at 87 meteorological stations distributed across Spain and Portugal for 1961–2019. Multidecadal variability analysis of both mean and gusts SWS confirms for the first time in the region the cessation of the stilling (decline of SWS) and a possible weak reversal phenomenon (increase in SWS) in the last decades, varying its onset from 1999 to 2018, depending on the season and variable. Different atmospheric circulation drivers are evaluated to explain the stilling and reversal phenomena. Among the chosen teleconnection indices, the Western Mediterranean Oscillation index (WeMOi) is the one that presents the highest (positive) correlation with SWS, although the correlation (negative) with the North Atlantic Oscillation index and the Mediterranean Oscillation index in winter is also statistically significant (p < 0.05). Results show that the interannual variability of the WeMOi could exert a strong influence on both the stilling and reversal phenomena, as it displays statistically significant negative trends for 1961–2010 and positive ones (non-significant) for 2010–2019, in agreement with the observed SWS trends. The use of the Jenkinson and Collison weather type classification reveals the marked influence of certain weather types in modulating SWS changes (i.e., west, northwest, and anticyclonic types). We proposed the increase in atmospheric thermal stability and the northward shift of the jet stream as principal causes of the stilling phenomenon in this region. Our results improve our understanding of wind changes, and highlight the importance of regional assessments to discern their socioeconomic and environmental impacts.

Wind power prediction based on EEMD-Tent-SSA-LS-SVM

To solve the wind power prediction problem, the Improved Sparrow Search Algorithm-Least Squares Support Vector Machine (ISSA-LS-SVM) prediction model based on chaotic sequences is proposed to improve the convergence accuracy and shorten the prediction time of the prediction model. Firstly, the problem in the historical data is decomposed using an ensemble empirical modal algorithm. Then, wind speed series prediction is performed using the LS-SVM model. Finally, the wind turbine output power prediction is performed. The results show that compared with LS-SVM, SSA-LS-SVM and Tent-SSA-LS-SVM models, the EEMD-ISSA-LS-SVM prediction model has improved the convergence precision of wind power output predictive model, which is significant for the subsequent realization of optimal power dispatch.

A multi-component hybrid system based on predictability recognition and modified multi-objective optimization for ultra-short-term onshore wind speed forecasting

The wind is a natural source of energy and wind energy occupies an important share in the global energy structure. Compared with offshore wind energy, the economy and availability of onshore wind energy make it dominate the wind energy industry at this stage. However, the instability and inconsistency of the ultra-short-term onshore wind speed will cause inefficiency with the turbines. Therefore, accurate onshore wind speed forecasts can help onshore wind farms enhance the efficiency of wind turbines and improve the accuracy of energy assessment. Most existing forecasting methods usually suffer from insufficient precision and high complexity. To improve these deficiencies, a multi-component hybrid onshore wind speed forecasting system based on predictability recognition framework, κ point modified multi-objective golden eagle optimizer, and weight hybrid kernel extreme learning machine is proposed in this study. According to the experiments of four onshore wind speed series collected from an onshore wind farm located in the US, the proposed system shows superior performances on one-step ahead and multi-step ahead forecast and it outperforms some typical methods. And the statistical significance of superior forecasting performance is fully demonstrated. Overall, the proposed forecasting system can offer a more reliable forecast for onshore wind speed.

Wind Speed Prediction for Offshore Sites Using a Clockwork Recurrent Network

Offshore sites show greater potential for wind energy utilization than most onshore sites. When planning an offshore wind power farm, the speed of offshore wind is used to estimate various operation parameters, such as the power output, extreme wind load, and fatigue load. Accurate speed prediction is crucial to the running of wind power farms and the security of smart grids. Unlike onshore wind, offshore wind has the characteristics of random, intermittent, and chaotic, which will cause the time series of wind speeds to have strong nonlinearity. It will bring greater difficulties to offshore wind speed predictions, which traditional recurrent neural networks cannot deal with for lacking in long-term dependency. An offshore wind speed prediction method is proposed by using a clockwork recurrent network (CWRNN). In a CWRNN model, the hidden layer is subdivided into several parts and each part is allocated a different clock speed. Under the mechanism, the long-term dependency of the recurrent neural network can be easily addressed, which can furthermore effectively solve the problem of strong nonlinearity in offshore speed winds. The experiments are performed by using the actual data of two different offshore sites located in the Caribbean Sea and one onshore site located in the interior of the United States, to verify the performance of the model. The results show that the prediction model achieves significant accuracy improvement.

Santa Ana Winds: Fractal-Based Analysis in a Semi-Arid Zone of Northern Mexico

A fractal analysis based on the time series of precipitation, temperature, pressure, relative humidity, and wind speed was performed for 16 weather stations located in the hydrographic basin of the Guadalupe River in Baja California, Mexico. Days on which the phenomenon known as Santa Ana winds occurs were identified based on the corresponding criteria of wind speed (≥4.5 m/s) and wind direction (between 0° and 90°). Subsequently, the time series was formed with data representing the days on which this phenomenon occurs in each of the analyzed weather stations. A time series was additionally formed from the days in which the Santa Ana winds condition does not occur. Hurst exponents and fractal dimension were estimated applying the rescaled range method to characterize the established time series in terms of characteristics of persistence, anti-persistence, or randomness along with the calculation of the climate predictability Index. This enabled the behavior and correlation analysis of the meteorological variables associated with Santa Ana winds occurrence. Finally, this type of research study is instrumental in understanding the regional dynamics of the climate in the basin, and allows us to establish a basis for developing models that can forecast the days of occurrence of the Santa Ana winds, in such a way that actions or measures can be taken to mitigate the negative consequences generated when said phenomenon occurs, such as fires and droughts.

A multi-layer extreme learning machine refined by sparrow search algorithm and weighted mean filter for short-term multi-step wind speed forecasting

With a rapidly growing wind power capacity, wind speed forecasting is of great importance for secure and economical operation of power systems. Nonetheless, due to the volatility of wind speed, its prediction has always been a challenging task. Although utilization of promising noise reduction techniques can mitigate this problem to some extent, wavelet threshold denoising methods (WTDs) suffer from their boundary effects in practice. To address this issue, weighted mean filtering (WMF) is adopted as an alternative of WTDs to reduce the information redundancy of wind speed time series. Furthermore, multi-layer extreme learning machine (MLELM) is combined with WMF to form an ensemble model for one-step and multi-step wind speed forecasting. Besides, as a novel swarm algorithm, the sparrow search algorithm (SSA), is utilized to improve the performance of MLELM. Finally, a data-driven ensemble model WMF-SSA-MLELM is proposed herein, which can be divided into three blocks, data preprocessing, optimization, and prediction. In the first block, WMF suppresses redundant noise in wind speed time series and makes it easier to extract essential features of wind speed. In the second block, SSA optimizes input weights and biases of hidden layers in the optimization stage. In the prediction block, MLELM with two hidden layers optimized by SSA provides the prediction of future wind speed by using denoised wind speed series obtained in the first two steps. The test results on four datasets demonstrate superior accuracy and efficiency of the proposed model WMF-SSA-MLELM, via comparison with all candidate models in both one-step and multi-step wind speed forecasting.

Ultra-short-term / short-term wind speed prediction based on improved singular spectrum analysis

Aiming at the problem that the wind speed data collected by wind farms are affected by many factors and easy to introduce noise information, a wind speed prediction method based on data decomposition of improved singular spectrum analysis (ISSA) is proposed. In this paper, the ISSA is used to decompose the wind speed sequence into a series of sub-sequences. Based on the singular spectrum analysis (SSA), the ISSA introduces the singular entropy to judge the noise components of the wind speed series and remove them. Then, the artificial neural network model is used to calculate and compare the prediction results of several data preprocessing decomposition methods using EMD, EEMD, CEEMD, ISSA and the prediction results without data preprocessing. Experimental results show that the proposed method can effectively improve the prediction accuracy of the artificial neural network, and also has higher prediction accuracy than the comparison method, which verifies the effectiveness of the ISSA.

Ultra-short-term wind speed forecasting based on EMD-VAR model and spatial correlation

Accurate wind speed forecasting is conducive to reduce the risk of power system from wind power uncertainty, which is of great significance to power system operation. However, it’s very difficult to achieve satisfactory results in wind speed forecasting due to the complex fluctuation characteristics of wind speed series. This study proposes a novel EMD-VAR wind speed forecasting model based on the wind speed data of multiple adjacent measuring points with high correlation. To achieve better experimental results with high accuracy and strong stability, multiple adjacent spatial sites are used to balance the information and variance of the forecasting model. Empirical mode decomposition (EMD) methods are utilized to remove the noise in the original data and multiple intrinsic mode function (IMF) components are obtained. For each IMF component, the corresponding vector autoregressive (VAR) model is established for the spatial groups. The final forecasting result is obtained by summarizing the forecasting results of all the IMF components. To validate the accuracy and stability of the proposed model, wind speed data sets in four seasons are used for experimental prediction. Experiments show that this method can effectively improve the accuracy and guarantee the reliability of wind speed forecasting in each season.

A novel hybrid approach based on variational heteroscedastic Gaussian process regression for multi-step ahead wind speed forecasting

Accurate wind speed forecasting is the key to safe and economic operation of electric power and energy systems. As a Bayesian nonparametric method, Gaussian process regression (GPR) has provided competitive forecasting results in recent years. However, conventional GPR model assumes that the noise obeys Gaussian distribution and the variance of the noise in the whole data set is a constant, which is not appropriate for some problems. Motivated by this, this study makes the first attempt to study the ability of the variational heteroscedastic GPR (VHGPR) model in wind speed forecasting. The Marginalized Variational (MV) approximation is employed to approximate the heteroscedastic Gaussian process in the VHGPR model. What’s more, the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is employed to transfer the nonstationary wind speed series into a certain number of subseries with more regularity such that the forecasting performance of VHGPR can be enhanced. The proposed method is compared with the other twelve benchmark models for 1- to 3- step ahead wind speed forecasting. Experiments results on four real-world datasets demonstrate that VHGPR with the CEEMDAN decomposition strategy is able to obtain better forecasting results for wind speed time series.

Short-term wind speed forecasting using deep reinforcement learning with improved multiple error correction approach

The safe and stable operation of wind power systems requires the support of wind speed prediction. To ensure the controllability and stability of smart grid dispatching, a novel hybrid model consisting of data-adaptive decomposition, reinforcement learning ensemble, and improved error correction is established for short-term wind speed forecasting. In decomposition module, empirical wavelet transform algorithm is used to adaptively disassemble and reconstruct the wind speed series. In ensemble module, Q-learning is utilized to integrate gated recurrent unit, bidirectional long short-term memory, and deep belief network. In error correction module, wavelet packet decomposition and outlier-robust extreme learning machine are combined to developing predictable components. An appropriate correction shrinkage rate is used to obtain the best correction effect. Ljung-Box Q-Test is utilized to judge the termination of the error correction iteration. Four real data are utilized to validate model performance in the case study. Experimental results show that: (a) The proposed hybrid model can accurately capture the changes of wind data. Taking 1-step prediction results as an example, the mean absolute errors for site #1, #2, #3, and #4 are 0.0829 m/s, 0.0661 m/s, 0.0906 m/s, and 0.0803 m/s, respectively; (b) Compared with several state-of-the-art models, the proposed model has the best prediction performance.

Short-term wind speed forecasting system based on multivariate time series and multi-objective optimization

Wind energy has attracted much attention because it is sustainable and renewable energy with a much smaller impact on the environment than fossil fuels. Accurate wind speed forecasting can pave the way to the goal of sustainable development. However, many papers only use wind speed series but ignore weather factors, which lead to poor forecasting results. In this paper, we propose an ensemble system that considers historical wind speed series and other weather factors to make wind speed forecasting. The proposed system integrates noise reduction, clustering, and multi-objective optimization. Firstly, the historical wind speed series are decomposed by complementary ensemble empirical mode decomposition (CEEMD). Secondly, the samples are clustered by self-organizing map (SOM) according to the weather factors. Finally, a regularized extreme learning machine (RELM) model is trained to forecast the wind speed for each cluster, and multi-objective grey wolf optimizer (MOGWO) is employed to optimize the parameters in the model. An empirical study using three datasets from the M2 tower of the national wind power technology center (NWTC) of the national renewable energy laboratory (NREL) illustrates that the proposed system performs better than other comparative models in terms of four performance indicators.

Multi-step wind speed forecasting based on secondary decomposition algorithm and optimized back propagation neural network

Improve the reliability of wind speed forecasting is a crucial task in wind power generation. Due to the stochastic and noise nature of wind, a preprocessing step is beneficial for wind speed series to get clean data. The decomposition technique is reported as the critical preprocessor to transform the unstable wind speed data into several regular components. This study proposes a hybrid forecasting system, which combines secondary decomposition algorithm and optimized back propagation (BP) neural network. For the decomposition part, the variational mode decomposition (VMD) is firstly used to extract the low-frequency part from the original wind data. Then the symplectic geometry mode decomposition (SGMD) decomposes the rest high-frequency part into clean and separate components. The BP algorithm is optimized by the differential evolution (DE) as the predictor in this study. Empirical studies with different comparison models are conducted on real wind speed data. The results affirm the competitive strength of the proposed combination strategy. And the proposed two-stage decomposition technique is applicable for nonlinear wind speed analysis.

A hybrid model for multi-step wind speed forecasting based on secondary decomposition, deep learning, and error correction algorithms

Accurate and stable wind speed forecasting is an essential means to ensure the safe and stable operation of wind power integration. Therefore, a new hybrid model was proposed to improve wind speed forecasting performance, consisting of data pre-processing, model forecasting, and error correction (EC). The specific modeling process is as follows: (a) A wind speed series was decomposed into a series of subseries with different frequencies utilizing the ensemble empirical mode decomposition (EEMD) method. Afterward, various subseries were divided into high-frequency components, intermediate-frequency components, and low-frequency components based on their sample entropies (SE). (b) Three frequency components were forecast by separately employing the hybrid model of convolutional neural network and long short-term memory network (CNN-LSTM), long short-term memory network (LSTM), and Elman neural network. (c) Subsequently, an error sequence was further forecast using CNN-LSTM. (d) Finally, three actual datasets were used to forecast the multi-step wind speed, and the forecasting performance of the proposed model was verified. The test results show that the forecasting performance of the proposed model is better than the other 13 models in three actual datasets.

A novel hybrid machine learning model for short-term wind speed prediction in inner Mongolia, China

Wind power is recognized as one of the most promising renewable and clean energy sources under the context of the increasing depletion of fossil fuels. The exact wind speed forecasting has great significance for the large-scale connection of wind farms with the power grid. In light of this, this paper contributes to establishing a novel hybrid model that can predict the future wind speed accurately. Firstly, the original wind speed time series is decomposed by the fast ensemble empirical mode decomposition into several sub-series that are further integrated by the runs test. The phase space reconstruction is used to dynamically choose each integrated sub-series' input and output vectors for the prediction model. Additionally, an improved whale optimization algorithm is exploited to optimize the weights and bias of the extreme learning machine. Finally, prediction results are obtained from the aggregation of each integrated sub-series prediction. To verify the accuracy and applicability of the proposed hybrid model, we apply several comparative models to conducted two case studies using different wind speed time series from Inner Mongolia that is Asia's largest gathering area of wind power farms. According to the experimental results, it can be concluded that the decomposition reduces the volatility and randomness of wind speed, and the runs test lowers the forecasting complexity. The phase space reconstruction can capture the chaotic property of wind speed series. The optimization for the whale optimization algorithm enhances its global and local optimization ability to further improve the performance of extreme learning machine. Overall, the proposed hybrid model can effectively capture the non-linear characteristics of wind speed series.

A novel combined model for wind speed prediction – Combination of linear model, shallow neural networks, and deep learning approaches

Accurate wind speed forecasting is increasingly essential for improving the operating efficiency of electric power systems. Numerous models have been proposed to obtain the accurate and stable wind speed forecasting results. However, previous proposed models are limited by single predictive model or cannot deal with complex nonlinear data characteristic, which resulted in poor and unstable prediction results. In this paper, a novel forecasting model that combines noise processing, statistical approaches, deep learning frameworks and multi-objective optimization algorithm is proposed. Multi-objective optimization algorithms can take advantage of the merits of benchmark prediction models to address nonlinear characteristics of wind speed series. The 10-min real wind speed data from three Sites in China are adopted for verifying the effectiveness of this proposed model. The experimental results of multi-step prediction show that the model achieves MAPE1-step = 2.2109%, MAPE2-step = 3.0309%, and MAPE3-step = 4.2536% at Site 1; MAPE1-step = 2.4586%, MAPE2-step = 3.2034%, and MAPE3-step = 4.6843% at Site 2; MAPE1-step = 2.3180%, MAPE2-step = 3.0846%, and MAPE3-step = 4.4193% at Site 3. Therefore, the forecasting performance of this model is excellent, and it is beneficial to the dispatching and planning of power grid.

Observation and modeling of Hurricane Maria for damage assessment

The human loss due to Hurricane Maria (H-Maria) in the month of September of 2017 was quantified to be more than 4500 casualties in the entire island of Puerto Rico, making it the most devastating storm in US history. Besides, H-Maria left a lasting impact on the Island as it brought to full collapse the electrical power grid rendering the Island entirely out of power for more than ten months. The aim of this work is to fill the gap of the hydro-meteorological processes of this relevant storm due to the limited observational data available. The synoptic observational record shows that the monthly average Sea Surface Temperature of 30 °C, with an anomaly of 0.5 °C as well as low vertical wind shear of 4–8 m/s fueled H-Maria. Simulated (WRF) time series of wind speed is in close resemblance compared with the limited data available from ocean buoys with a simulated and observed peak wind speed of 30 m/s on the southern coast of the Island. The total rainfall for the event was simulated to peak at 762 mm (observed 965 mm) at the center of the Island and was validated with post-hurricane National Weather Service (NWS) rainfall with a Normalized Root Mean Square Error (RMSE) of 0.2. The orographic effects are simulated, reflecting enhancement of the rainfall at high altitudes in the central mountains of the Island. As an example of damage assessment, the risk of failure of the electrical power towers as a function of wind speed and soil saturation is simulated using statistical models for the entire Island, which results in higher risks of failure at the Northwest and center of the Island. These validated results of the storm can also be used as an input for other analysis such as hydrological models to geo-locate regions for risks due to flooding.

Nonstationary wind speed data reconstruction based on secondary correction of statistical characteristics

Data reconstruction is usually used for lost data recovery and data supplement in structural health monitoring (SHM) system. In this paper, a data reconstruction algorithm is proposed for the entire time series data missing at dispersed and consecutive multiple observation points. Kriging-based sequence interpolation (KSI) is adopted for primary data reconstruction and probability density function (PDF) reconstruction. In terms of dispersed data missing, the reconstructed PDF is of great accuracy, which is used to calculate the exact standard deviation to correct the primary reconstructed data and obtain satisfactory performance. Regarding the consecutive data missing, the first correction is insufficient due to the missing of too much information. In addition, it is found to exist a fuzzy mapping relation between the exact standard deviation and extreme values of wind speed series. Thus, the quantile regression combined with deep neural network is proposed for probability reconstruction of the extreme values. The point estimations are adopted to the secondary correction for the reconstructed data with first correction. The effectiveness and validity of the secondary correction strategy have been testified by structural dynamic response analysis. The extreme value of the displacement response under the reconstructed data with the secondary correction is closer to that under the actual wind data, compared with the reconstructed data with first correction.