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

For large wind farms with grid-connected power generation systems, power prediction is very important. A sparrow search algorithm is proposed for optimizing the wind power prediction in BP neural network to improve the accuracy and stability of wind power prediction. Tent chaotic mapping gets used to increase the species diversity of the sparrow search algorithm and improve the ability of the algorithm to step out of local optimization and global search. The improved SSA-BP algorithm and traditional BP neural network method are used to predict the wind power. The prediction results and errors of the two methods are compared. The comparative analysis of the actual measured data and model prediction data shows that the SSA-BP method can better track the trend of the actual wind speed data, reduce the error and improve the prediction accuracy.

A generalized data preprocessing method is proposed in this paper to reduce the amount of outliers among historical data and further improve the power prediction accuracy. Historical data of wind farms are fit with an S-shape curve via Linear Regression Model. Based on this statistical curve, outliers can be identified considering different fitting error. Furthermore, the expansion of wind farm is identified through the number of outliers. Then a selection method for the allowed maximum fitting errors is recommended. The presented method has been integrated into the prediction system in Inner Mongolia of China with 36 farms. The actual application shows that the wind farm power prediction accuracy has been improved by at least 28% with this model. It is noteworthy that the proposed preprocessing method is just based on statistical analysis of historical data and thus compatible with various wind power prediction methods.

A short-term wind power prediction method is proposed in this paper with experimental results obtained from a wind farm located in Northeast China. In order to improve the accuracy of the prediction method using a traditional back-propagation (BP) neural network algorithm, the improved grey wolf optimization (IGWO) algorithm has been adopted to optimize its parameters. The performance of the proposed method has been evaluated by experiments. First, the features of the wind farm are described to show the fundamental information of the experiments. A single turbine with rated power of 1500 kW and power generation coefficient of 2.74 in the wind farm was introduced to show the technical details of the turbines. Original wind power data of the whole farm were preprocessed by using the quartile method to remove the abnormal data points. Then, the retained wind power data were predicted and analysed by using the proposed IGWO–BP algorithm. Analysis of the results proves the practicability and efficiency of the prediction model. Results show that the average accuracy of prediction is ~11% greater than the traditional BP method. In this way, the proposed wind power prediction method can be adopted to improve the accuracy of prediction and to ensure the effective utilization of wind energy. A short-term wind power prediction method is designed and tested with experimental results obtained from a wind farm located in Northeast China. In order to improve the accuracy of the prediction method, the improved grey wolf optimization algorithm has been adopted to optimize its parameters.

Short-term wind power prediction based on multidimensional data cleaning and feature reconfiguration

Accurate wind power prediction is of great significance for ensuring the stable operation of power systems. Although the rapid development of deep learning technology has significantly improved the accuracy of wind power prediction, existing methods often lack sufficient capability for information interaction among variables when processing multivariate time series data, and struggle to capture long-range dependencies in longer sequences, which limits their predictive performance. To address this, this paper proposes a novel model named Efficient Feature Aggregation based on Mamba (EFAMamba) for multistep wind power prediction. This model introduces Mamba to more effectively capture long-term dependencies in time series. The EFA module aggregates all variable sequences into a global representation, which is then dispatched and fused with each variable sequence representation, significantly promoting information interaction among variables. The entire framework of EFAMamba can comprehensively consider historical data of multiple variables as well as future meteorological variable information, making it particularly suitable for multi-step wind power prediction. Extensive experiments conducted on real wind farm dataset have validated that the proposed EFAMamba demonstrates significant advantages over current state-of-the-art (SOTA) methods in terms of prediction accuracy for multi-step forecasting tasks, particularly excelling under longer prediction horizons. Specifically, when predicting 72 hours ahead, the NMAE, NRMSE, and QR are improved by 9.96%, 3.77%, and 1.39%, respectively, demonstrating the effectiveness of the proposed EFAMamba in wind power prediction. The source code and datasets are available at https://github.com/USTCAI4EEE/EFAMamba.

With the gradual warming of the global climate, wind energy, as one of the representatives of clean energy, has been rapidly developed in countries worldwide. However, wind power generation has strong intermittency and variability, constantly impacting the stability of the power grid. In order to improve the accuracy of wind power prediction, this paper proposes a BP neural network model optimized by genetic algorithm (GA) for wind power prediction. This model uses GA to optimize the initial weights and thresholds of the BP neural network to improve the computational accuracy and convergence speed of the model, effectively avoiding the problem of a single BP neural network model easily falling into local minima and the slow training speed of GA algorithm. By preprocessing and normalizing meteorological data such as weather, temperature, and wind direction, a wind power prediction model is constructed, and the prediction results show that the model effectively improves the computational accuracy and convergence speed compared to the traditional BP neural network model. This result proves the feasibility of using GA to optimize BP neural network for wind power prediction in this paper.

This paper presents a method to improve the short-term wind power prediction at a given turbine using information from numerical weather prediction and from multiple observation points, which correspond to the locations of nearby turbines at a particular wind farm site. The prediction of wind power is achieved in two stages; in the first stage wind speed is predicted using our proposed method. In the second stage, the wind speed to output power conversion is accomplished using power curve model. The proposed wind power prediction method is tested using real measurements and NWP data from one of the wind farm sites in Australia. The performance is compared with the persistence and Grey predictor model in terms of Mean Absolute Error and Root Mean Square Error.

Wind power plays a more and more important part in today’s electricity market with the increasing need of clean energy. How to make better use of wind power is attracting great attention due to the inherent uncertainty caused by extreme complexity of weather systems. In order to address difficulty, we introduced a time-series prediction method called FEDformer to the field of wind power forecasting, and made some adjustment so that the FEDformer could perform better in this field. We tested this novel model on a real-world wind farm dataset, and compared it with other traditional machine learning methods. Results show that the proposed method for wind power prediction has better accuracy and greater potential for engineering applications. It is effective to apply transformer-based model to the field of wind power prediction.

Accurate short-term forecasting of wind farm power generation is of great significance to economic development and stable operation of power system. In order to improve the accuracy of wind power prediction, this paper presents a short-term wind power prediction method based on empirical mode decomposition and deep learning. We firstly decompose historical data into two-dimensional tensors by using empirical mode decomposition, and then extract the local features of data by using convolutional neural network. The features are transmitted to the bidirectional long short-term memory for prediction. The measured wind power data form wind farm in Xinjiang are used for verification. We use root mean square error and mean absolute error as evaluation indexes to test the algorithm. Results of the experiment show that the reconstructed two-dimensional tensor is beneficial to the ability of convolutional neural network to extract local features, and the prediction accuracy of our method is higher than that of other traditional prediction algorithms.

Wind power prediction accuracy is closely related to the quality of wind power data. However, most of traditional prediction methods do not consider the cybersecurity of raw data. So, it would be important to study the prediction accuracy under false data injection attack (one type of cybersecurity attacks) and its influence on the economy and security of power system. In this paper, very short-term forecasting approach of wind power generation is studied for choosing the optimal prediction model. Different scenarios of false data injection attack are then simulated upon wind power data. Wind power prediction accuracy after being affected by false data injection is observed. The robustness of wind power prediction model is evaluated. Finally, wind power prediction information is applied to day-ahead unit commitment to evaluate the impact of cybersecurity of wind power forecasting on power system economy and security.

Wind power predication is significant for the safety and stability of power system operation. Due to the power forecasting single model or combination model without considering the information fusion and the influence of forecasting error on power dispatching, forecasting results are not accurate enough. In the paper, a novel combination model for wind power forecasting with error evaluation parameter based on cross entropy theory is proposed. Wind power prediction is seen as information fusion, and the weights of different wind power prediction methods is set according to their cross levels. Besides, an error evaluation parameter is also proposed to estimate the different influences of positive and negative error on power scheduling. Application results in an actual wind farm show that the cross levels of different wind power prediction methods can be identified effectively and prediction precision is improved by the proposed method.

Accurate wind power prediction is of great significance for grid stability and renewable energy integration. Addressing the challenge of effectively integrating physical mechanisms with data-driven methods in wind power prediction, this paper innovatively proposes a two-stage deep learning prediction framework incorporating physics-guided fusion and distribution constraints, aiming to improve the prediction accuracy and physical authenticity of individual wind turbines. In the first stage, we construct a baseline model based on multi-branch multilayer perceptrons (MLP) that eschews traditional attempts to accurately reconstruct complex three-dimensional spatiotemporal wind fields, instead directly learning the power conversion characteristics of wind turbines under specific meteorological conditions from historical operational data, namely the power coefficient (Cp). This data-driven Cp fitting method provides a physically interpretable and robust benchmark for power prediction. In the second stage, targeting the prediction residuals from the baseline model, we design a bidirectional long short-term memory network (BiLSTM) for refined correction. The core innovation of this stage lies in introducing Maximum Mean Discrepancy (MMD) as a regularization term to constrain the predicted wind speed-power joint probability distribution. This constraint enforces the model-generated power predictions to remain statistically consistent with historical real data distributions, effectively preventing the model from producing predictions that deviate from physical reality, significantly enhancing the model’s generalization capability and reliability. Experimental results demonstrate that compared to traditional methods, the proposed method achieves significant improvements in Mean Absolute Error, Root Mean Square Error, and other metrics, validating the effectiveness of physical constraints in improving prediction accuracy.

Accurate and timely wind power forecast is difficult because of the volatility and intermittency of wind energy, which are strongly influenced by meteorological conditions. In order to solve this problem, this research suggests a data-driven wind power prediction technique based on Elman neural networks and an enhanced generative adversarial network (GAN). First, outliers in the supervisory control and data acquisition (SCADA) power and wind speed data are found by combining the methods of density clustering and regression threshold truncation. To guarantee the temporal continuity of wind power data with the original characteristics, minority class samples are then generated using an enhanced GAN and added to the dataset. In order to obtain an accurate wind power prediction, the balanced data is then fed into a prediction model that combines the ABC algorithm and Elman neural networks. The fitness value is the mean squared error (MSE) of the training data, and the optimization objective is the connection weights of the Elman neural network. Results from experiments show that the suggested model outperforms earlier state-of-the-art techniques and significantly increases wind power prediction accuracy. It also has the advantages of high stability and quick convergence speed, and it can capture the long-term dependencies of wind power data.

Regional wind power prediction (WPP) is conducive to achieving scientific dispatch of the power system as well as improving its reliability. Data mining based on a large number of samples is an effective way to improve the accuracy of WPP. A new regional power forecasting method in the context of data mining is proposed in the article. First, a previously described regional WPP method, single feature similarity matching (SFSM) method, is analyzed and a new multifeature similarity matching (MFSM) method is proposed. Second, the optimization and analysis processes of four key parameters in the MFSM method are proposed and the impact of each key parameter on prediction error and the applicability of the method under different regional scales is analyzed. Finally, the new method and the optimization and analysis processes are verified by wind farm clusters, composed of 52 wind farms in northeast China, and the new method is compared with the conventional method, such as additive method, statistical up-scaling method, and the SFSM method. The results show that the improved MFSM method is superior to the conventional prediction method, whereby the root mean square error (RMSE) of 0 to 2 h is reduced by 3.65% and the RMSE of 0 to 4 h is reduced by 1.56%. The new method has been proved to be an effective WPP method for wind farm clusters and has potential industrial application prospects.

Accurate wind power prediction can reduce the negative impact of wind power on power grid and reduce the operation cost of power system. Wind power forecasting is of great significance in spot trading, but the instability of wind power series brings difficulties to forecasting. With the rapid development of deep learning technology, prediction algorithm based on deep learning has been widely used in wind power prediction. The long short-term memory (LSTM) network performs well in a variety of prediction scenarios, but there is still room for improvement. In order to improve the prediction accuracy of LSTM in wind power prediction, this paper adopts a new gating mechanism in LSTM network, which can improve the context information modeling ability of LSTM network. The experimental results based on real data show that the proposed improved LSTM prediction model can improve the prediction accuracy of LSTM prediction model in ultra-short term wind power prediction scenarios.