Short-term Wind Power Forecasting Research Articles

Short-Term Forecasting of Wind Power Based on Error Traceability and Numerical Weather Prediction Wind Speed Correction

Short-Term Forecasting of Wind Power Based on Error Traceability and Numerical Weather Prediction Wind Speed Correction

Short-term Wind Power Forecasts based on VMD-KPCA and CFSBOA-BiLSTM

Accurate prediction of wind power is of great significance to reduce the impact of large-scale wind power connection and improve the security and stability of power grid. A short-term wind power forecasting method based on VMD-KPCA and CFSBOA-BiLSTM is proposed. Firstly, the key environmental factors restricting the volatility of wind power are fully considered, and the variational mode decomposition method is used to decompose the wind power and meteorological factors. Secondly, the kernel principal component analysis method is used to reduce the dimension and construct the sub-sequence characteristic data set. Finally, the multi-strategy improved butterfly algorithm is used to optimize the BiLSTM network hyperparameters and establish the CFSBOA-BiLSTM prediction model. The simulation analysis uses the measured data of a wind farm in northwest China. The experimental results show that the combined forecasting model can fully mine the hidden information of the data and improve the accuracy of short-term wind power forecasting. Compared with the BiLSTM model, RMSE, MAE and MAPE decreased by 60.37%, 66.6% and 67.59%, respectively.

A novel Bagged-CNN architecture for short-term wind power forecasting

ABSTRACT Wind power forecasting plays a pivotal role in ensuring the dependable integration of wind energy into power grids. This research introduces a novel Bagged-CNN architecture tailored for short-term wind power forecasting. This architecture harnesses the combined strengths of the Bagging ensemble technique and Convolutional Neural Networks (CNN) to enhance prediction accuracy and robustness. Our approach demonstrates superior forecasting capabilities by exploiting the spatial and temporal features captured by CNNs. To validate our proposed architecture, we have conducted a comprehensive evaluation across three distinct wind farms, each characterized by unique geographical and climatic conditions. We systematically compared the performance of our Bagged-CNN model against several standalone machine-learning algorithms commonly employed in wind power forecasting. The hyperparameters of our Bagged-CNN model are meticulously optimized using the GRID search algorithm, which systematically explores a predefined parameter space. Notably, our predictions shed light on the substantial impact of training data period length on wind power prediction accuracy. Our research underscores the potential of coupling Bagging and CNN techniques for short-term wind power forecasting. The Bagged-CNN architecture, as proposed, serves as a robust framework for capturing intricate spatial and temporal patterns in wind power data, thereby facilitating more precise and dependable predictions.

Wavelet decomposition and neural networks: a potent combination for short term wind speed and power forecasting

The forecast of wind speed and the power produced from wind farms has been a challenge for a long time and continues to be so. This work introduces a method that we label as Wavelet Decomposition-Neural Networks (WDNN) that combines wavelet decomposition principles and deep learning. By merging the strengths of signal processing and machine learning, this approach aims to address the aforementioned challenge. Treating wind speed and power as signals, the wavelet decomposition part of the model transforms these inputs, as appropriate, into a set of features that the neural network part of the model can ingest to output accurate forecasts. WDNN is unconstrained by the shape, layout, or number of turbines in a wind farm. We test our WDNN methods using three large datasets, with multiple years of data and hundreds of turbines, and compare it against other state-of-the-art methods. It’s very short-term forecast, like 1-h ahead, can outperform some deep learning models by as much as 30%. This shows that wavelet decomposition and neural network are a potent combination for advancing the quality of short-term wind forecasting.

Adaptive temporal transformer method for short-term wind power forecasting considering shift in time series distribution

In wind power prediction, the input probability distributions in the different sub-periods are shifted owing to the strong randomness of the input features, such as wind speed and direction. This may violate the assumption for machine learning that the training and test data meet the condition of being independent and identically distributed, resulting in an insufficient generalization ability of the prediction model that is trained with the training data and applied to unknown test data. To address this problem, this study proposes an adaptive temporal transformer method for short-term wind power forecasting. First, a temporal transformer model with a gate recurrent unit and multi-head attention layers was used to extract the short- and long-term temporal information of the multiple input variables. Then, an adaptive learning strategy consisting of two stages—temporal distribution characterization and temporal distribution matching—was developed to explore the common knowledge hidden in each sub-period. The case results for an actual wind farm in northwest China showed that the proposed method could effectively weaken the adverse effects of the shifts in time series distribution on forecasting and improve the accuracy and generalization of short-term wind power prediction.

Short-term wind power forecasting based on dual attention mechanism and gated recurrent unit neural network

Accurate wind power forecasting is essential for both optimal grid scheduling and the massive absorption of wind power into the grid. However, the continuous changes in the contribution of various meteorological features to the forecasting of wind power output under different time or weather conditions, and the overlapping of wind power sequence cycles, make forecasting challenging. To address these problems, a short-term wind power forecasting model is established that integrates a gated recurrent unit (GRU) network with a dual attention mechanism (DAM). To compute the contributions of different features in real time, historical wind power data and meteorological information are first extracted using a feature attention mechanism (FAM). The feature sequences collected by the FAM are then used by the GRU network for preliminary forecasting. Subsequently, one-dimensional convolution employing several distinct convolution kernels is used to filter the GRU outputs. In addition, a multi-head time attention mechanism (MHTAM) is proposed and a Gaussian bias is introduced to assign different weights to different time steps of each modality. The final forecast results are produced by combining the outputs of the MHTAM. The results of the simulation experiment show that for 5-h, 10-h, and 20-h short-term wind power forecasting, the established DAM-GRU model performs better than comparative models on the basis of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), R-squared (R2), Square sum error (SSE), Mean absolute percentile error (MAPE), and Relative root mean square error (RRMSE) index.

Wind Power Short-Term Time-Series Prediction Using an Ensemble of Neural Networks

Short-term wind power forecasting has difficult problems due to the very large variety of speeds of the wind, which is a key factor in producing energy. From the point of view of the whole country, an important problem is predicting the total impact of wind power’s contribution to the country’s energy demands for succeeding days. Accordingly, efficient planning of classical power sources may be made for the next day. This paper will investigate this direction of research. Based on historical data, a few neural network predictors will be combined into an ensemble that is responsible for the next day’s wind power generation. The problem is difficult since wind farms are distributed in large regions of the country, where different wind conditions exist. Moreover, the information on wind speed is not available. This paper proposes and compares different structures of an ensemble combined from three neural networks. The best accuracy has been obtained with the application of an MLP combiner. The results of numerical experiments have shown a significant reduction in prediction errors compared to the naïve approach. The improvement in results with this naïve solution is close to two in the one-day-ahead prediction task.

Least Squares-based Optimal Reconciliation Method for Hierarchical Forecasts of Wind Power Generation

Wind power generation is hierarchically organized and can be aggregated/disaggregated based on geography and electrical network structure. Forecasts are required at all levels of such a hierarchy for different operational requirements, which are also known as “hierarchical forecasts”. In this paper, we propose a new method to produce hierarchical wind power forecasts which performs better than the conventional top-down or bottom-up method. At first, wind power generation at all levels of the hierarchy is independently forecasted using state-of-the-art techniques. Then, a least squares regression model is used to optimally reconcile these base forecasts. Adjusted forecasts are as close as possible to base forecasts but also satisfy the requirement of the aggregate consistency (i.e., the lower-level forecasts add up exactly to the higher-level forecasts). Case studies validate the effectiveness of the proposed approach for very short-term wind power forecasts. The accuracy improvement of reconciled forecasts over benchmarks is confirmed at all levels of the hierarchy, regardless of forecast horizons and hierarchical structures.

Short-Term Wind Power Forecast Based on Continuous Conditional Random Field

Short-Term Wind Power Forecast Based on Continuous Conditional Random Field

Relative evaluation of probabilistic methods for spatio-temporal wind forecasting

Short-term wind power forecasting has become a de facto tool to better facilitate the integration of such renewable energy resources into modern power grids. Instead of point predictors, which produce single-value predictions for the expected power, probabilistic forecasts predict probability distributions over the expected power output or associated confidence intervals. In this study, three different parametric and non-parametric methods for uncertainty modelling in wind power forecasting were studied, namely quantile regression (QR), variational inference and a maximum likelihood estimation (MLE) method. Johnson’s SU distribution was studied as a novel candidate for modelling wind power, which is a transformed normal distribution that exhibits both skew and heavy tails. This was one of the first studies to provide a thorough investigation of Johnson’s SU distribution for uncertainty modelling in a complex deep learning framework for wind forecasting. It was found that Johsnon’s SU likelihood and QR-based models significantly outperformed models using Gaussian likelihoods, based on a range of quantitative metrics to evaluate probability distributions and qualitative investigation of produced forecasts. Variational inference models using Johnson’s SU likelihoods performed remarkably well, with near-perfect calibration and higher precision than models using any of the other methods for uncertainty modelling, as evaluated through the pinball loss, Average Coverage Error (ACE) and Prediction Interval Coverage Percentage (PICP) metric. With the superior performance of Johnson’s SU likelihood models, the study mainly contributes to the literature by introducing another candidate distribution for probabilistic wind forecasting, which is analytical, unbounded and easy to integrate into modern deep learning frameworks.

Short-term wind power forecasting based on multivariate/multi-step LSTM with temporal feature attention mechanism

Precision enhancement for short-term wind power forecasting can alleviate negative impact of the forecasting results on wind power generation. Due to complexities and nonlinearities among factors and facets in wind power, it is essential to achieve reliable and stable power generation via the long short-term memory (LSTM) forecasting. To this purpose, multi-task temporal feature attention (MTTFA) based LSTM, namely MTTFA-LSTM, is proposed for multivariate/multi-step wind power forecasting with historical power and meteorological data, in which task-sharing and task-specifying layers are designed for task co-features extracting and task specifics discriminating, respectively. More specifically, in the task-sharing layer, multi-dimensional inputs are fed into LSTM to extract long-term trends, while in the task-specifying layer, one-dimensional convolution operations extract temporal features hidden in each and all time steps. Furthermore, an attention mechanism is adopted to adaptively tune weights for temporal features. Finally, the proposed model is leveraged to cope with different short-term wind power forecasting (SWPF) problems based on the national renewable energy laboratory’s (NREL) wind power data. Simulation results show that the proposed MTTFA-LSTM achieves persistent excellent forecasting accuracy, comparing its backbone STL model, TFA-LSTM as well as the benchmark MTL models in the same setting, which indicate that the complex and non-linear interdependencies among multi-dimensional data can be well depicted by the proposed model.

Multistep short-term wind power forecasting model based on secondary decomposition, the kernel principal component analysis, an enhanced arithmetic optimization algorithm, and error correction

Wind power forecasting can effectively improve the energy utilization efficiency of a power system and ensure its stable operation. In this study, a novel hybrid multistep prediction model, including the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), variational mode decomposition (VMD), the kernel principal component analysis (KPCA), an enhanced arithmetic optimization algorithm (ENAOA), a bidirectional long short-term memory (BILSTM) neural network, and error correction, was designed for short-term wind power forecasting. First, the collected original wind power data were decomposed into multiple intrinsic mode functions (IMFs) through a secondary decomposition composed of the CEEMDAN and VMD, which eliminated the interactions between different components to achieve denoising. Second, the KPCA was adopted to reduce the dimensionality of the multiple IMFs by extracting the principal components, effectively reducing the complexity of the multidimensional IMF data and improving the forecasting efficiency of the proposed prediction model. Subsequently, an ENAOA was proposed based on the Sobol sequence, adaptive T-distribution, and random walk strategy to optimize the BILSTM parameters. Finally, multiple preprocessed components were predicted by the optimized BILSTM, after which error correction was performed to obtain the final prediction results, which could further reduce the forecast error of the designed prediction model. Based on two sets of data collected from a wind farm in northwest China, the simulation results of 1-step, 4-step, 7-step, and 10-step forecasting revealed that compared with other incomplete models, the various algorithms adopted in the hybrid forecasting model reduced the prediction errors to different degrees, significantly enhanced the wind power prediction performance, and validated the effectiveness and feasibility of the proposed model.

Short-Term Wind Speed and Power Forecasting for Smart City Power Grid With a Hybrid Machine Learning Framework

To address foreseeable challenges during the penetration of wind energy into the power grid, including accurate wind power forecasting and smart power generation scheduling, this study proposes a novel short-term wind speed forecasting model, named EMD-KM-SXL, which is based on empirical mode decomposition (EMD), K-means clustering and machine learning, and a new two-stage short-term wind power forecasting model based on wind speed forecasting and wind power curve modeling. The former wind speed forecasting model regards historical wind speed observations as model inputs, and the latter power forecasting model utilizes knowledge augmentation, introducing wind power conversion relationship, environmental factors as well as wind power system status parameters. In the proposed wind speed forecasting model, three machine learning models, including support vector regressor, XGBoost regressor, and Lasso regressor, are employed to forecast three types of frequency components that are generated via EMD and K-means clustering. Then, the wind power curve model is utilized to compute potential outpower based on the predicted wind speed, which is regarded as the first stage of the proposed wind power forecasting model. In the second stage, environmental factors and wind power system status parameters are introduced and an artificial neural network model, considering preliminary predicted power, environmental factors, and wind power system status parameters as model inputs, is built to make final power prediction. Computational results show that proposed models achieve the best performance in terms of wind speed and power forecasting over different forecasting horizons ranging from 10 to 40 minutes, compared with benchmarking methods.

A Multichannel-Based CNN and GRU Method for Short-Term Wind Power Prediction

Incorporating wind energy on a large scale into power systems presents challenges for the operation and control of the grid. To enhance the safety of power grid operation, accurate short-term forecasting of wind power is necessary, as it minimizes the impact of randomness. Considering the uncertainty and prediction issues associated with wind power, this paper introduces a CNN–GRU ultra-short-term wind power prediction model. This model relies on multichannel signals, including data such as wind speed, wind direction, climate conditions, and historical power outputs collected from wind farms. These data types contribute to the formation of a comprehensive multichannel signal for wind power. Following that, the CNN method extracts both global and partial features from these signals. Concurrently, features are extracted from past power outputs based on their time series. These features are then combined with the ones obtained from the convolution process. Subsequently, these combined features are input into a fully connected network. This step is crucial for blending multichannel information and generating forecast results. To validate the model, it was tested using data from a wind farm located in a specific region of Sichuan Province. According to our experimental results, the model demonstrates a high level of accuracy in computation and robust generalization ability.

A hybrid model based on discrete wavelet transform (DWT) and bidirectional recurrent neural networks for wind speed prediction

Wind speed is the main driver of wind power output, but its inherent fluctuations and deviations present significant challenges for power system security and power quality. Accurate short-term wind power forecasting is necessary to ensure the stability and seamless integration of wind energy into the grid, thereby bolstering grid reliability, optimizing power flow management, and minimizing disruptions, ultimately contributing to a more sustainable and efficient energy ecosystem. Non-stationarity is a major challenge in analyzing wind speed data, and change-point detection is essential for optimal resource allocation. This paper addresses the issue of short-term wind power forecasting for stable and effective wind energy system operation. To predict non-stationary data and detect change points, non-stationary data must first be transformed into stationary data. Discrete wavelet transform (DWT) is used to decompose wind speed traces into low- and high-frequency components for more accurate predictions using deep learning algorithms. The proposed approach utilizes a combination of DWT and recurrent neural network (RNN) models, including Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Unit (BiGRU), complementing each other to efficiently predict short-term and long-term dependencies in wind speed data and achieve the most accurate prediction results. Experimental results substantiate the consistent superiority of the proposed approach over the baseline method, showcasing a significant enhancement in prediction accuracy, with improvements spanning from 11.36% to 16.63% for MAPE, 16.67%–26.47% for RMSE, and 21.74%–30.77% for MAE.

Short-term wind power forecasting by bidirectional attention mechanism LSTM and its probability interval prediction by sliding-window KDE

Deterministic point wind power forecasting (DP-WPF) and its probability interval prediction (PIP) are indispensable to short-term peak alleviation and frequency regulation in power systems with large-scale wind power injection. To improve short-term DP-WPF by long short-term memory (LSTM), a horizontal/vertical bidirectional feature attention (BFA) based LSTM model is proposed. More specifically, the BFA-LSTM model has three parts: first, multivariate time series are fed into LSTM to extract long-short-term temporal features; second, the LSTM outputs are processed horizontally as well as vertically for retrieving step-wise/multistep-wise temporal features, respectively, namely, in the bidirectional attention sense; third, both the horizontal and vertical attention weights are adaptively adjusted according to the feature importance in DP-WPF. Cases comparison shows that the suggested modeling is stably superior to most common counterparts. To address PIP by kernel density estimation (KDE), sliding-window KDE is leveraged for probability analysis. More precisely, probability density functions (PDF) and probability intervals are estimated with sliding-window samples, which are non-parametric operations and involve finitely many local samples. Superior performances of PIP by sliding-window KDE at different confidence levels indicate that the sliding-window PDF approach is highly effective in contrast to those with all samples.

A Short-Term Wind Power Forecasting Model Based on 3D Convolutional Neural Network–Gated Recurrent Unit

Enhancing the accuracy of short-term wind power forecasting can be effectively achieved by considering the spatial–temporal correlation among neighboring wind turbines. In this study, we propose a short-term wind power forecasting model based on 3D CNN-GRU. First, the wind power data and meteorological data of 24 surrounding turbines around the target turbine are reconstructed into a three-dimensional matrix and inputted into the 3D CNN and GRU encoders to extract their spatial–temporal features. Then, the power predictions for different forecasting horizons are outputted through the GRU decoder and fully connected layers. Finally, experimental results on the SDWPT datasets show that our proposed model significantly improves the prediction accuracy compared to BPNN, GRU, and 1D CNN-GRU models. The results show that the 3D CNN-GRU model performs optimally. For a forecasting horizon of 10 min, the average reductions in RMSE and MAE on the validation set are about 10% and 11%, respectively, with an average improvement of about 1% in R. For a forecasting horizon of 120 min, the average reductions in RMSE and MAE on the validation set are about 6% and 8%, respectively, with an average improvement of about 14% in R.

Wind power forecasting based on SCINet, reversible instance normalization, and knowledge distillation

Wind energy plays a crucial role as a clean energy source in the electricity system. The unpredictability of wind power makes it more challenging to put into use in comparison to thermal power generation. Accurate wind power prediction algorithms are of great importance for allocation and deployment of wind power. In this paper, a novel time-series forecasting model, SCINet, is used for short-term wind power forecasting and achieves high forecasting accuracy. Furthermore, the addition of reversible instance normalization (RevIN) to SCINet effectively alleviates the shift problem that arises in time series forecasting tasks. This enhancement further improves the model's forecasting ability. Finally, this paper uses knowledge distillation to get a small model that could speed up the computing and save memory resources. The source code is available at https://github.com/raspnew/WPF.git.

Probabilistic wind power forecasting for newly-built wind farms based on multi-task Gaussian process method

The accurate training of a wind power forecasting (WPF) model for a newly built wind farm is difficult because of limited historical data. This study established a multitask learning architecture wherein the WPF in different wind farms represents an independent task. Subsequently, a novel short-term WPF model based on a multitask learning architecture was proposed. In this model, a multitask Gaussian process is used to capture the intertask conjunction, which contributed to the training of each task. The proposed methodological framework employs dependencies from other tasks wherein older wind farms contain substantial historical data to enhance the performance of tasks in which there is a newly built wind farm. Several numerical experiments were conducted using datasets from seven independent wind farms in Australia. The results show that the proposed scheme not only obtains improved point forecasting results but also produces better probabilistic forecasting results, thus demonstrating the superiority of the proposed method.

Reconstruction of Unsteady Wind Field Based on CFD and Reduced-Order Model

Short-term wind power forecasting is crucial for updating the wind power trading strategy, equipment protection and control regulation. To solve the difficulty surrounding the instability of the statistical model and the time-consuming nature of the physical model in short-term wind power forecasting, two innovative wind field reconstruction methods combining CFD and a reduced-order model were developed. In this study, POD and Tucker decomposition were employed to obtain the spatial–temporal information correlation of 2D and 3D wind fields, and their inverse processes were combined with sparse sensing to reconstruct multi-dimensional unsteady wind fields. Simulation and detailed discussion were performed to verify the practicability of the proposed algorithms. The simulation results indicate that the wind speed distributions could be reconstructed with reasonably high accuracy (where the absolute velocity relative error was less than 0.8%) using 20 sensors (which only accounted for 0.04% of the total data in the 3D wind field) based on the proposed algorithms. The factors influencing the results of reconstruction were systematically analyzed, including all-time steps, the number of basis vectors and 4-mode dimensions, the diversity of CFD databases, and the reconstruction time. The results indicated that the reconstruction time could be shortened to the time interval of data acquisition to synchronize data acquisition with wind field reconstruction, which is of great significance in the reconstruction of unsteady wind fields. Although there are still many studies to be carried out to achieve short-term predictions, both unsteady reconstruction methods proposed in this paper enable a new direction for short-term wind field prediction.