Research on short-term wind power forecasting method based on incomplete data

To meet the increasing wind power forecasting (WPF) demands of newly built wind farms without historical data, physical WPF methods are widely used. The computational fluid dynamics (CFD) pre-calculated flow fields (CPFF)-based WPF is a promising physical approach, which can balance well the competing demands of computational efficiency and accuracy. To enhance its adaptability for wind farms in complex terrain, a WPF method combining wind turbine clustering with CPFF is first proposed where the wind turbines in the wind farm are clustered and a forecasting is undertaken for each cluster. K-means, hierarchical agglomerative and spectral analysis methods are used to establish the wind turbine clustering models. The Silhouette Coefficient, Calinski-Harabaz index and within-between index are proposed as criteria to evaluate the effectiveness of the established clustering models. Based on different clustering methods and schemes, various clustering databases are built for clustering pre-calculated CFD (CPCC)-based short-term WPF. For the wind farm case studied, clustering evaluation criteria show that hierarchical agglomerative clustering has reasonable results, spectral clustering is better and K-means gives the best performance. The WPF results produced by different clustering databases also prove the effectiveness of the three evaluation criteria in turn. The newly developed CPCC model has a much higher WPF accuracy than the CPFF model without using clustering techniques, both on temporal and spatial scales. The research provides supports for both the development and improvement of short-term physical WPF systems.

In this study, wind power forecasting is performed for a Wind Power Plant (WPP) with an installed capacity of 135 MW in Turkey. The ruggedness index (RIX) of the terrain where WPP was installed is analyzed with Wind Atlas Analysis and Application Program (WAsP). According to the obtained RIX value, the terrain of WPP is found to be complex. Due to the complexity of the terrain, wind power forecasting becomes difficult. To deal with this problem, a forecasting method with fast, accurate, and high performance is needed. Therefore, Extreme Learning Machine (ELM) based method is proposed for wind power forecasting in this study. Electrical and meteorological measurements are obtained from WPP for the application of the proposed method. These measurements are provided with high quality measuring devices. Also, Global Positioning System (GPS) time synchronization is used to prevent lags between measurements. The wind speed, wind direction, and wind power data of 1-year period are obtained from WPP. These data are used to compare the proposed method with a classical Artificial Neural Network (ANN) based method in terms of two, three and four hours-ahead wind power forecast performances. In the forecast studies performed for all data related to 2, 3, and 4-hours ahead, Normalized Root Mean Square Error (NRMSE) values of ELM are obtained as 7.01, 10.12, and 12.06, respectively, while these values are found as 8.19, 12.18, and 13.09 for ANN. In addition, the values of Correlation Coefficients (R) of the proposed forecast method results regarding 2, 3, and 4-hours ahead are 0.96588, 0.93528, and 0.88984, respectively. The R values related to ANN are observed as 0.95421, 0.91373, and 0.87576, respectively. According to the obtained results, it is observed that ELM has better performance features than classic method under all forecast conditions and it is clearly seen that ELM has by far short training time than other one.

Accurate forecasting is the key factor in promoting wind power consumption and improving the stable operation of power systems. A short-term wind power forecasting (WPF) and uncertainty analysis method based on whale optimization algorithm (WOA), least squares support vector machine (LSSVM), and nonparametric kernel density estimation (NPKDE) was proposed in this paper. The advantages of WOA (fast convergence speed and high convergence accuracy) were used to optimize the penalty factor and kernel function width of the LSSVM model, and the calculation speed and forecasting accuracy of the LSSVM model were improved. The training sample set is classified according to the wind speed interval, and the WOA-LSSVM forecasting model is trained by subclass after classification to further improve the accuracy of short-term WPF. The NPKDE method is used to accurately calculate the probability density distribution characteristics of the forecasting error of wind power, and the confidence interval of the WPF is accurately calculated based on the probability density distribution characteristics. The calculation results show that the forecasting accuracy of the WOA-LSSVM model is higher than those of the LSSVM, long short-term memory (LSTM), and particle swarm optimization and least squares support vector machine (PSO-LSSVM) models, and the forecasting accuracy of the WOA-LSSVM model can be further improved after classifying the training sample set. The coverage of the confidence intervals in different time scales is higher than the corresponding confidence level, indicating that the NPKDE method can accurately describe the probability density distribution characteristics of the WPF errors.

Accurate short-term wind power forecasting has a significant impact to economic dispatch, which ensures the efficiency and smooth operation of power system. In this paper, a hybrid wind power forecasting method based on a typical algorithm of feedforward neural network, convolutional neural network (CNN) and a time circulation neural network, and long short-term memory network (LSTM) models is proposed. It aims to improve the accuracy of ultra-short-term wind power forecasting.

The intermittency, uncontrollability, and variability of wind power affect the economical operation and reliable delivery of the power system. To ensure the smooth integration of wind power into the grid, an accurate wind power forecast is essential. In this paper, we propose a short-term wind power prediction model based on decomposition and reconstruction to improve the accuracy of wind power forecasts. This model has the following characteristics: (i) wind power time series are decomposition into multiple components through the application of a decomposition method that combines fully integrated empirical mode decomposition with adaptive noise algorithm; (ii) the components are clustered using K-means clustering based on dynamic time warping. According to the similarity of the complexity and lag of components, they are divided into three classes; (iii) three different prediction models, including seasonal autoregressive integrated moving average model (SARIMA), long short-term memory network (LSTM) and Bi-directional long short-term memory network (BiLSTM), predict three types of components respectively. Finally, to illustrate the capability of the model, we compare its performance with three typical models. The results demonstrate that the proposed method exceeds the baseline models in regard to prediction performance.

We propose a novel statistical wind power forecast framework, which leverages the spatio-temporal correlation in wind speed and direction data among geographically dispersed wind farms. Critical assessment of the performance of spatio-temporal wind power forecast is performed using realistic wind farm data from West Texas. It is shown that spatio-temporal wind forecast models are numerically efficient approaches to improving forecast quality. By reducing uncertainties in near-term wind power forecasts, the overall cost benefits on system dispatch can be quantified. We integrate the improved forecast with an advanced robust look-ahead dispatch framework. This integrated forecast and economic dispatch framework is tested in a modified IEEE RTS 24-bus system. Numerical simulation suggests that the overall generation cost can be reduced by up to 6% using a robust look-ahead dispatch coupled with spatio-temporal wind forecast as compared with persistent wind forecast models.

Accurate forecasting methods for wind and solar power are important for power systems because of their potential to improve the economic and environmental performance. For this purpose, an ensemble neural network framework composed of LSTM, SVM, BP neural network, and ELM is proposed for wind and solar power forecasting in China. Three common methods for improving the prediction accuracy were adopted. First, unstable wind and solar power time series are decomposed into smooth subsequences by VMD, which reduces the undesirable effects caused by the volatility of the original series. Then, based on the decomposed subsequences, four basic models that are optimized based on the EOSSA algorithm are used to forecast wind and solar power. Finally, the prediction results of ENN were reconstructed by weighting the prediction results of the four models. The proposed ENN model was compared to nine state-of-the-art prediction models for wind and solar power forecasting. The results showed that the ENN model had the lowest MAPE, MAE, MSE, and RMSE for both wind and solar power forecasting. These comparison results also show that the ENN model not only has the best prediction accuracy, but also the most reliable prediction performance.

This article suggests the application of multiresolution analysis by Wavelet Transform—WT and Echo State Networks—ESN for the development of tools capable of providing wind speed and power generation forecasting. The models were developed to forecast the hourly mean wind speeds, which are applied to the wind turbine’s power curve to obtain wind power forecasts with horizons ranging from 1 to 24 h ahead, for three different locations of the Brazilian Northeast. The average improvement of Normalized Mean Absolute Error—NMAE for the first six, twelve, eighteen and twenty-four hourly power generation forecasts obtained by using the models proposed in this article were 70.87%, 71.99%, 67.77% and 58.52%, respectively. These results of improvements in relation to the Persistence Model—PM are among the best published results to date for wind power forecasting. The adopted methodology was adequate, assuring statistically reliable forecasts. When comparing the performance of fully-connected feedforward Artificial Neural Networks—ANN and ESN, it was observed that both are powerful time series forecasting tools, but the ESN proved to be more suited for wind power forecasting.

Wind power is one of the most efficient renewable resources without emissions. Nonetheless, it is difficult to exactly forecast wind power generation given historical power and wind speed information, the failure of which may cost the risk of large-scale outages. This article takes a close look at the artificial recurrent neural network framework in the application of wind power forecasting. More intelligent mechanisms using attention to capture spatial-temporal patterns within historical data are emphasized in this work and are shown to be state-of-the-art for short-term wind power forecasting. Our experiments at a wind farm in southeast Australia using only the historical wind power generation and wind speed records from ambient weather stations show that, e.g., 7.4750% in mean absolute error (MAE) and 0.3345 in the coefficient of variation in the root mean squared error (CV-RMSE) for half-hour-ahead prediction. To interpret how the three models under consideration—the long- and short-term time-series network (LSTNet), the temporal pattern attention-based long short-term memory (TPA-LSTM) and the dual-stage attention-based recurrent neural network (DA-RNN)—work, we visualize and analyze the details of the models so that further improvement can be made by combining the advantageous components of the models.

A new multistage short-term wind power forecast model using decomposition and artificial intelligence methods

The increasing penetration of renewable energy, and wind power in particular, requires accurate short-term forecasting to ensure grid stability, reduce operational uncertainty, and facilitate large-scale integration of intermittent resources. This study evaluates Transformer-based architectures for wind power forecasting using hourly generation data from Spain (2020–2024). Time series were segmented into input windows of 12, 24, and 36 h, and multiple model configurations were systematically tested. For benchmarking, LSTM and GRU models were trained under identical protocols. The results show that the Transformer consistently outperformed recurrent baselines across all horizons. The best configuration, using a 36 h input sequence with moderate dimensionality and shallow depth, achieved an RMSE of 370.71 MW, MAE of 258.77 MW, and MAPE of 4.92%, reducing error by a significant margin compared to LSTM and GRU models, whose best performances reached RMSEs above 395 MW and MAPEs above 5.7%. Beyond predictive accuracy, attention maps revealed that the Transformer effectively captured short-term fluctuations while also attending to longer-range dependencies, offering a transparent mechanism for interpreting the contribution of historical information to forecasts. These findings demonstrate the superior performance of Transformer-based models in short-term wind power forecasting, underscoring their capacity to deliver more accurate and interpretable predictions that support the reliable integration of renewable energy into modern power systems.

Reliable wind power forecasting is essential for maintaining the safe and stable operation of power systems with high renewable energy penetration. This study proposes a short-term wind power forecasting model based on decomposition–optimization–prediction, integrating complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN), the improved football team training algorithm (IFTTA), and the bidirectional long short-term memory network model (BiLSTM). CEEMDAN is employed to decompose the non-stationary wind power sequence into relatively stable intrinsic mode functions (IMFs), thereby separating multi-scale fluctuation features. The IFTTA incorporates a dynamic probability allocation strategy and an adaptive parameter adjustment mechanism, which contributes to a better balance between global exploration and local exploitation. After optimizing the hyperparameters of BiLSTM using IFTTA, the prediction performance significantly improved. Validations were conducted on three datasets from Xinjiang, Ningxia, and Inner Mongolia, China, each containing 1440 samples (1152 for training and 288 for testing). Comparisons with the benchmark forecasting model demonstrate that the pro-posed model reduces the mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) by at least 25.29%, 29.62%, and 20.66%, respectively. Correspondingly, the coefficient of determination (R2) was improved by at least 0.0069. This model provides an effective solution for short-term wind power prediction in practical engineering.

Accurate and stable wind power forecasting is an inevitable requirement for efficient use of renewable energy. This paper proposes a short-term wind power combination forecasting method based on mode decomposition and feature extraction. The original wind speed time series is decomposed into multiple subsequences by the ensemble empirical mode decomposition (EEMD). The generalized mutual information (GMI) is used to extract the optimal input feature set of each subsequence. The prediction values of each subsequence are obtained based on the least squares support vector machine (LSSVM) model, and then the final prediction results are obtained by combining them. Finally, the paper proves that the proposed method can predict short-term wind power more accurately and stably by setting up the contrast experiment.

Using enhanced crow search algorithm optimization-extreme learning machine model to forecast short-term wind power

This article presents the weather and power data files from renewable sources used to solve the economic dispatch problem of a microgrid that operates in the isolated and grid-connected modes. Methodology is used in the research article “Management of an island and grid-connected microgrid using hybrid economic model predictive control with weather data” (Silva et al., 2020). Automatic stations located in the Brazil's south and northeast furnished the weather data (global horizontal irradiance, temperature, and wind speed). A script generates files containing weather forecasts from one-day ahead using the geographical coordinates of the weather stations. Hybrid models, characterized by real and binary variables, use the weather forecasting data to calculate the photovoltaic and wind power forecasts. A microgrid management algorithm uses these forecasts to solve the optimal economic dispatch problem. This data-in-brief paper presents five datasets for each weather station: (i) Weather dataset downloaded from the website of the National Meteorological Institute, (ii) Weather research and forecasting (WRF) dataset derived from the raw data generated by the weather research and forecasting model, (iii) Weather dataset that joins the forecasted data with the measured data in a single file, (iv) Handled dataset that treats some gaps in the weather dataset and converts it to other formats, (v) Files containing only the temperature, global horizontal irradiance, and wind speed data, (vi) Files containing the measured and forecasted wind and solar power.