Short-term Wind Speed Prediction Research Articles

Short-term Wind Speed Prediction with a Two-layer Attention-based LSTM

Wind speed prediction is of great importance because it affects the efficiency and stability of power systems with a high proportion of wind power. Temporal-spatial wind speed features contain rich information; however, their use to predict wind speed remains one of the most challenging and less studied areas. This paper investigates the problem of predicting wind speeds for multiple sites using temporal and spatial features and proposes a novel two-layer attention-based long short-term memory (LSTM), termed 2Attn-LSTM, a unified framework of encoder and decoder mechanisms to handle temporal-spatial wind speed data. To eliminate the unevenness of the original wind speed, we initially decompose the preprocessing data into IMF components by variational mode decomposition (VMD). Then, it encodes the spatial features of IMF components at the bottom of the model and decodes the temporal features to obtain each component's predicted value on the second layer. Finally, we obtain the ultimate prediction value after denormalization and superposition. We have performed extensive experiments for short-term predictions on real-world data, demonstrating that 2Attn-LSTM outperforms the four baseline methods. It is worth pointing out that the presented 2Atts-LSTM is a general model suitable for other spatial-temporal features.

Tačnost predvidljivosti brzine vetra sa visinom korišćenjem rekurentnih neuronskih mreža

Accurate prediction of wind speed in future time domain is critical for wind power integration into the grid. Wind speed is usually measured at lower heights while the hub heights of modern wind turbines are much higher in the range of 80-120m. This study attempts to better understand the predictability of wind speed with height. To achieve this, wind data was collected using Laser Illuminated Detection and Ranging (LiDAR) system at 20m, 40m, 50m, 60m, 80m, 100m, 120m, 140m, 160m, and 180m heights. This hourly averaged data is used for training and testing a Recurrent Neural Network (RNN) for the prediction of wind speed for each of the future 12 hours, using 48 previous values. Detailed analyses of short-term wind speed prediction at different heights and future hours show that wind speed is predicted more accurately at higher heights.For example, the mean absolute percent error decreases from 0.19 to 0.16as the height increase from 20m to 180m, respectively for the 12 th future hour prediction. The performance of the proposed method is compared with Multilayer Perceptron (MLP) method. Results show that RNN performed better than MLP for most of the cases presented here at the future 6th hour.

Short-term wind speed prediction based on CEEMDAN-SE-improved PIO-GRNN model

In this paper, a short-term wind speed prediction model, called CEEMDAN-SE-Improved PIO-GRNN, is proposed to optimize the accuracy of the short-term wind speed forecast. This model is established on account of the optimized General Regression Neural Network (GRNN) method optimized by three algorithms, which are Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Sample Entropy (SE), and Pigeon Inspired Optimization (PIO), separately. Firstly, decomposing the original wind speed sequences into several subsequences with different complexity by CEEMDAN. Then, the complexity of each subsequence is judged by SE and the similar subsequences are combined into a new sequence to reduce the scale of calculation. Afterwards, the GRNN model optimized by improved PIO is used to predict each new sequence. Finally, the predicted results are superposed as the eventual predicted value. Implementing the prediction for the wind speed data of a wind field in north China within 30 days by applying the different prediction models, namely, GRNN, CEEMDAN-GRNN, Improved PIO-GRNN, and CEEMDAN-SE-Improved PIO-GRNN which are proposed in this paper. Comparing the prediction curves of different models with the fitting curve of the actual wind speed shows that the optimal fitting effect and minimum error value are included in CEEMDAN-SE-Improved PIO-GRNN model. Specifically, the values of mean squared error (MSE), mean absolute error (MAE) and weighted mean absolute percentage error (WMAPE) separately decrease by 0.6222, 0.3334, and 8.5766%, which compare with the single prediction model GRNN. Meanwhile, diebold-mariano (DM) test shows that the prediction ability of the two models is significantly different. The above statements indicate the proposed model does great advance in the precision of short-term wind speed prediction.

Short-term wind speed prediction using Extended Kalman filter and machine learning

Wind speed prediction could play an important role in improving the performance of wind turbine control and condition monitoring. For example, by predicting or forecasting the upcoming wind in advance, fluctuations in wind power output in above rated wind speed could be reduced without causing an increase in pitch activity, and anomalies such as an extreme gust could be detected before it reaches the wind turbine, allowing appropriate control actions to take place to minimise any potential damage that could be incurred by the anomalies. A novel wind speed prediction scheme is presented in this paper that comprises mainly two stages, estimation and prediction. Estimation is first carried out using an Extended Kalman filter, which is designed based on a 3 dimensional wind field model and a nonlinear rotor model. Prediction is subsequently performed in two steps, extrapolation and machine learning. The wind speed prediction scheme is tested using data obtained from a high-fidelity aeroelastic model.

Multistep short-term wind speed prediction using nonlinear auto-regressive neural network with exogenous variable selection

Precise wind speed prediction is a key factor in many energy applications, especially when wind energy is integrated with power grids. However, because of the intermittent and nonstationary nature of wind speed, modeling and predicting it is a challenge. In addition, using uncorrelated multivariate variables as exogenous input variables often adversely impacts the performance of prediction models. In this paper, we present a multistep short-term wind speed prediction using multivariate exogenous input variables. We implement different variable selection methods to select the best set of variables that significantly improve the performance of prediction models. We evaluate the performance of eight transfer learning methods, four shallow neural networks (NNs), and the persistence method on predicting the future values of wind speed using ultrashort-term, short-term, and multistep time horizons. We performed the evaluation over two-year high-sampled wind speed data averaged at 10-minute intervals. Results show that Nonlinear Auto-Regressive Exogenous (NARX) model outperformed all other methods, achieving an average mean absolute error (MAE) and root mean square error (RMSE) of 0.2205 and 0.3405 for multistep predictions, respectively. Despite the lower performance of the transfer learning methods (i.e., 0.43 and 0.58 for MAE and RMSE, respectively), it is believed that results could be further improved with a better enhancement of the feature selection and model parameters.

Short-term wind speed predicting framework based on EEMD-GA-LSTM method under large scaled wind history

Accurate short-term wind speed prediction is of great significance for early warning and regulation of wind farms. At present, the scale of wind speed time-history data is increasing, and its time resolution is also becoming higher. Traditional machine learning models cannot effectively capture and utilize nonlinear features from the large scaled dataset and this, not only increases the difficulty of model building, but also reduces the prediction accuracy. To overcome such challenges, a machine learning based framework involving data-mining method was proposed in this paper. To begin with, a powerful signal decomposition technique (ensemble empirical mode decomposition) was used to divide the original wind sequence into several intrinsic mode functions to form a potential feature set. Then, a more appropriate sub-feature set together with the corresponding machine learning model were automatically generated through an iteration process. Such process was constructed through a coupled algorithm using the binary coded searching method known as the genetic algorithm and the advanced recurrent neural network with long short term memory unit. The analytical results show that, when compared with the traditional mainstream models, the strategy of using the sequences provided by the signal decomposition technology as the input features can significantly improve the prediction accuracy. On the other hand, faced with the high-dimensional feature set generated from the big data, the selected sub-feature set can not only perform a large dimension reduction, but also further improve the prediction accuracy up to 28.33% in terms of different kinds of evaluation criteria. Therefore, there is a potential application of the proposed method on more accurate short-term wind speed prediction under a considerable dataset of wind history.

Preliminary Research of Chaotic Characteristics and Prediction of Short-Term Wind Speed Time Series

Short-term wind speed prediction has its special significance in wind power industry. However, due to the characteristics of the wind system itself, it is not easy to predict the short-term wind speed accurately. In order to solve the problem, this paper studies the chaotic characteristics and prediction of short-term wind speed time series. The short-term wind speed data at four time scales are collected as the research object. The predictability of short-term wind speed time series is determined by the Hurst exponent. The chaotic characteristics of short-time wind speed at different time scales are analyzed by the 0–1 test method for chaos and the maximum Lyapunov exponent method. The results show that the short-term wind speed time series has chaotic characteristics at different time scales. The phase-space reconstruction technology is introduced; delay time is determined by the C–C method; embedding dimension is obtained by the G–P method. Echo state network is improved to suppress the influence of input noise on prediction performance. At the same time, an improved grey Wolf optimization algorithm is proposed to optimize the parameters of reserve pool of the echo state network. The results of a case study show that, compared with state-of-the-art methods, the proposed prediction method improves the prediction accuracy and reduces the predictive errors.

A Markov model for short term wind speed prediction by integrating the wind acceleration information

Wind speed prediction is an important research topic in the wind industry and many algorithms have been proposed to fulfill the prediction tasks. By reviewing the existing methods, one can find that the supplemental information, such as acceleration and turbulence intensity, that can be indirectly derived from wind speed is still less considered in the prediction models. To make a better utilization of these indirect information and future enhance the prediction accuracies, this paper proposes a novel Markov prediction model to integrate the wind acceleration. The proposes method first encodes the wind speed sequence into a discrete state sequence based on a 2D codebook that associates with the joint distribution of speed and acceleration. The discrete state sequence is then utilized to compute the state Transition Probably Matrix (TPM). The TPM governs the underlying state transition mechanisms in the Markov chain (or the state sequence) and serves as key to predict the states into the future time horizon. Lastly, the predicted states sequences are decoded into wind speed and the prediction uncertainty can be described as predictive distributions. The proposed method holds several advantages such as enhanced prediction accuracy and excellent flexibility to encode the supplemental information into the prediction model. The effectiveness of the proposed method is verified in the case studies by benchmarking with existing methods.

Very Short-term Wind Speed Prediction Using Geostatistical Kriging with External Trend

The prediction of wind speed plays a key role in the determination of the optimal reserve capacity for different time instants and the reduction of the standby cost of wind power grid. Short-term wind speed prediction can reduce the impact of wind power grid and improve the control of wind turbines. The geostatistics kriging as an unbiased estimation method applied to predict the wind speed in a two-dimensional vertical rectangular grid. The wind data were as a geographical variable considered. We firstly used exploratory methods to examine the spatial variability of wind speed and its dependence on geographical variables. The covariance and variogram models took into account in analysing the spatial structure of wind speed at three different instants of time. Then the product model adopted to fit wind data due to the variogram. The accuracy of the method was assessed through a cross validation. The prediction results were comparable with the observed values. The prediction error was about ±0.15 and the correlation coefficient was about 0.85. This study demonstrated that the kriging method could be success fully used in the short-time wind speed prediction in space and time direction.

Longitudinal coherence and short-term wind speed prediction based on a nacelle-mounted Doppler lidar

The spatial structure of turbulence in atmospheric boundary layer flows is highly relevant to wind energy. In particular, wind turbine control strategies based on inflow preview measurements require knowledge of the longitudinal evolution of turbulent flow as it approaches the rotor. These upstream measurements are usually obtained with nacelle-mounted wind lidars. In contrast to traditional in situ anemometry, lidars collect measurements within a probe volume which varies in size depending on the technology of the commercial system being used. Here, we address two issues related to the use of wind lidar to measure the incoming flow to a wind turbine: (i) whether existing longitudinal coherence models can be used to predict flow at the rotor, based on measurements performed at a distance away from the rotor; and (ii) what effect probe-volume averaging has on the inflow predictions. These two questions are critical to the design and implementation of robust wind turbine control strategies. To address these questions, we perform field measurements and large-eddy simulations to determine which incoming flow structures can be readily predicted with existing coherence models, and which require additional corrections to account for lidar volumetric averaging effects. Results reveal that the wind turbine induction zone has a negligible impact on the longitudinal coherence and first-order turbulence quantities, such as the standard deviation of velocity fluctuations. However, the phase of the signal, from which advection time periods of the turbulent structures are derived, is affected by the rotor blockage effect.

Short-Term Wind Speed Prediction Based on Principal Component Analysis and LSTM

An accurate prediction of wind speed is crucial for the economic and resilient operation of power systems with a high penetration level of wind power. Meteorological information such as temperature, humidity, air pressure, and wind level has a significant influence on wind speed, which makes it difficult to predict wind speed accurately. This paper proposes a wind speed prediction method through an effective combination of principal component analysis (PCA) and long short-term memory (LSTM) network. Firstly, PCA is employed to reduce the dimensions of the original multidimensional meteorological data which affect the wind speed. Further, differential evolution (DE) algorithm is presented to optimize the learning rate, number of hidden layer nodes, and batch size of the LSTM network. Finally, the reduced feature data from PCA and the wind speed data are merged together as an input to the LSTM network for wind speed prediction. In order to show the merits of the proposed method, several prevailing prediction methods, such as Gaussian process regression (GPR), support vector regression (SVR), recurrent neural network (RNN), and other forecasting techniques, are introduced for comparative purposes. Numerical results show that the proposed method performs best in prediction accuracy.

Short-term wind speed prediction using time varying filter-based empirical mode decomposition and group method of data handling-based hybrid model

The realization of precise and reliable short-term wind speed prediction is extremely essential to wind power development, especially for its integration into traditional grid system. For this purpose, this study develops a novel forecasting method based on time varying filter-based empirical mode decomposition, auto-regressive integrated moving average model and group method of data handling-based hybrid model. This method mainly contains four individual steps for grasping the major behavioral characteristics of wind speed data. The first step adopts time varying filter-based empirical mode decomposition to handle the nonlinearity and nonstationarity of the raw wind speed data by decomposing them into a number of subseries with more stability and regularity. Then, auto-regressive integrated moving average model is applied to depict the linear characteristic hidden in the data. For the above modeling errors (i.e., the nonlinear residuals), the third step employs three nonlinear models with different action mechanisms (i.e., least square support vector machine, genetic programming algorithm and spatio-temporal radial basis function neural network) to systematically capture their complex nonlinear features. Finally, group method of data handling neural network is utilized to combine these nonlinear models and perform the selective prediction, where the involved models and their weights could be determined automatically. Four groups of the measured wind speed datasets with two different time intervals are used to assess the performance of the proposed method. The experimental results indicate it outperforms the other compared models and may have great potential for the practical application in power system.

Improved regularized extreme learning machine short-term wind speed prediction based on gray correlation analysis

Accurate wind speed prediction is of great significance to the stable operation of the power grid when large-scale wind power is connected to the grid. This article proposes a new multi-step wind speed combination prediction model based on gray correlation analysis. First, the gray correlation analysis is performed on wind speed–related attributes, the more relevant attribute factors are selected as the input set of the prediction model, and the regularized extreme learning machine improved by the cuckoo optimization algorithm is used to perform multi-step prediction of wind speed. Then, an error self-tuning model is established to further improve the prediction accuracy. Finally, the measured results of different wind farms and seasons are selected to simulate the prediction effect of the proposed model, and the prediction accuracy and generalization ability of the proposed model are verified through comparative analysis.

Research on Offshore Short-term Wind Speed Prediction Based on the CSA Modeling Improved by Random Algorithm

Aiming at improving the prediction accuracy of offshore short-term wind speed, a model based on random cuckoo search algorithm (RCSA) and artificial neural network (ANN) was proposed. Firstly, RCSA was obtained by introducing a random factor to improve CSA, and then a RCSA-ANN model for predicting offshore short-term wind speed was established. Secondly, a wind tower was built in Luchao Port, Shanghai, offshore meteorological data were measured, and the training of the model was carried out. Finally, the precision of RCSA-ANN model was verified by comparison and analysis with BP-ANN and CSA-ANN models. The results show that the improved CSA method is simple, reliable and effective, which solves the problem that the algorithm is easy to fall into local optimum. The average error of RCSA-ANN model is not only lower than that of BP-ANN model, but also much lower than that of CSA-ANN model, and the prediction accuracy of these three models decreases in turn. RCSA-ANN model has high prediction accuracy and can precisely predict fluctuating wind speed sequences, and it also has good application potential.

Turbine-specific short-term wind speed forecasting considering within-farm wind field dependencies and fluctuations

The unprecedented scale and sophistication of wind turbine technologies call for wind forecasts of high spatial resolution, i.e. turbine-tailored forecasts, to inform several operational decisions at the turbine level. Towards that, this paper is concerned with leveraging the hub-height measurements collected from a fleet of turbines on a farm to make turbine-specific short-term wind speed and power predictions. We find that the wind propagation across a dense grid of turbines induces strong spatial and temporal dependencies in the within-farm wind field, but also gives rise to high-frequency high-magnitude fluctuations which may compromise the predictive accuracy of several data-driven forecasting methods. To capture both aspects, we propose to model the total variability in the within-farm wind speed field as a combination of two independent stochastic process terms. The first term reconstructs and extrapolates the wind speed field by learning the complex spatio-temporal dependence structure using hub-height turbine-level data. The second term accounts for high-frequency high-magnitude fluctuations that are not informed by near-term spatio-temporal dependencies. The two terms are coupled to make probabilistic wind speed forecasts at each turbine, which are then translated into turbine-specific power predictions via wind power curves. Evaluation on more than 3,000,000 data points from a wind farm dataset provides a strong empirical evidence in favor of the proposed method’s forecasting accuracy. On average, our proposed method achieves 9% accuracy improvement relative to persistence forecasts, and 7–9% relative to a set of widely recognized forecasting methods such as autoregressive-based models and Gaussian Processes.

A new combined model based on multi-objective salp swarm optimization for wind speed forecasting

Wind energy as the representative renewable energy sources attracted the global attention and wind power plays a significant role in power system. Thus, wind speed forecasting is highly critical in wind power grid management. The short-term wind speed prediction can effectively support power grid-management to reduce wind curtailments. In the past, lots of researches had often considered how to enhance the accuracy or stability in short wind speed forecasting. Nevertheless, just focus on one criterion is the inability to build an effective predictive system. In this paper, a novel combined forecasting system was proposed and effectively applied to address the issue of wind speed prediction while obtaining high precision and strong stability simultaneously at the same time. Four ANNs (artificial neural networks) were combined by the optimal weighting coefficients determined by MSSO (multi-objective salp swarm optimizer) in this system and data decomposition and denoising are included in the data preprocessing stage. The multi-objective optimization algorithm overcomes the weakness of the single-objective optimization algorithm that can only achieve one criterion. It can simultaneously optimize accuracy and stability. The 10-minute wind speed data of three data sets of Penglai, China were selected for multi-step forecasting to evaluate the effectiveness of the proposed combined model. And experimental results show that the proposed model not only achieves excellent precision and stability but also outperforms other proposed combined models.

Multi-Step Short-Term Wind Speed Prediction Using a Residual Dilated Causal Convolutional Network with Nonlinear Attention

Wind energy is the most used renewable energy worldwide second only to hydropower. However, the stochastic nature of wind speed makes it harder for wind farms to manage the future power production and maintenance schedules efficiently. Many wind speed prediction models exist that focus on advance neural networks and/or preprocessing techniques to improve the accuracy. Since most of these models require a large amount of historic wind data and are validated using the data split method, the application to real-world scenarios cannot be determined. In this paper, we present a multi-step univariate prediction model for wind speed data inspired by the residual U-net architecture of the convolutional neural network (CNN). We propose a residual dilated causal convolutional neural network (Res-DCCNN) with nonlinear attention for multi-step-ahead wind speed forecasting. Our model can outperform long-term short-term memory networks (LSTM), gated recurrent units (GRU), and Res-DCCNN using sliding window validation techniques for 50-step-ahead wind speed prediction. We tested the performance of the proposed model on six real-world wind speed datasets with different probability distributions to confirm its effectiveness, and using several error metrics, we demonstrated that our proposed model was robust, precise, and applicable to real-world cases.

A Hybrid Nonlinear Forecasting Strategy for Short-Term Wind Speed

The ability to predict wind speeds is very important for the security and stability of wind farms and power system operations. Wind speeds typically vary slowly over time, which makes them difficult to forecast. In this study, a hybrid nonlinear estimation approach combining Gaussian process (GP) and unscented Kalman filter (UKF) is proposed to predict dynamic changes of wind speed and improve forecasting accuracy. The proposed approach can provide both point and interval predictions for wind speed. Firstly, the GP method is established as the nonlinear transition function of a state space model, and the covariance obtained from the GP predictive model is used as the process noise. Secondly, UKF is used to solve the state space model and update the initial prediction of short-term wind speed. The proposed hybrid approach can adjust dynamically in conjunction with the distribution changes. In order to evaluate the performance of the proposed hybrid approach, the persistence model, GP model, autoregressive (AR) model, and AR integrated with Kalman filter (KF) model are used to predict the results for comparison. Taking two wind farms in China and the National Renewable Energy Laboratory (NREL) database as the experimental data, the results show that the proposed hybrid approach is suitable for wind speed predictions, and that it can increase forecasting accuracy.

Short-term wind speed prediction based on MLP and NARX network models

The article aims to predict the wind speed by two artificial neural network’s models. The first model is a multilayer perceptron (MLP) treated by back-propagation algorithm and the second one is a recurrent neuron network type, processed by the NARX model. The two models having the same Network’s structure, which they are composed by 4 Inputs layers (Wind Speed, Pressure Temperature and Humidity), an intermediate layer defined by 20 neurons and an activation function, as well as a single output layer characterized by wind speed and a linear function. NARX shows the best results with a regression coefficient R = 0.984 et RMSE = 0.314.

Short-term forecasting of wind speed using time division ensemble of hierarchical deep neural networks

Wind power in tropical countries like India has great potential as a major source of green energy. However, in order to do proper energy provisioning, there is a need to forecast and estimate the wind speed at windmill locations with an actionable lead-time. Forecasting wind speed at station level is a big challenge using dynamical models as it gives only macro level information. Therefore, use of statistical models is preferably adopted for this purpose. With the recent phenomenal growth of applications in artificial intelligence (AI), it is also possible to use data-driven models based on AI, especially deep learning for short-term forecasting of wind speed. In this paper, we have proposed a novel ensemble forecasting methodology using the long short-term memory (LSTM) model, which is a deep learning approach for time series data analysis. The capability of this approach has been demonstrated using wind speed data obtained from two meteorological stations located at New Delhi in North India and at Bengaluru in South India. We have used the ensemble methodology in two different modes; one is the averaging pooling and other is by using a hierarchical LSTM. The simulations using these models have been validated against the true observations at station scale. The ensemble forecasting method has shown promising results for 3-h early wind speed prediction at both the locations. The results are also compared with two classical statistical methods namely autoregressive and persistence models and two state-of-the-art data-driven models namely support vector machine (SVM) and extreme learning machine (ELM). The capability of the proposed method is demonstrated through various error matrices and found to have better performance. We believe that the proposed method has the potential to improve the short-term wind speed prediction capability at station level.