Short-Term Wind Speed Forecast With Low Loss of Information Based on Feature Generation of OSVD

Abstract

Improving the accuracy of wind speed forecast can reduce the randomness and uncertainty of the wind power output and effectively improve a system's wind power accommodation. However, the highdimensional historical wind speed information should be taken into account in the wind speed forecast, which increases the complexity of the model and reduces the efficiency and accuracy of a forecast. Feature selection by the Filter method can effectively reduce the feature dimension, but losing all the information of low-importance features. Although the feature reduction can retain the partial information of all features, it causes the loss of the partial information of high-importance features. In order to reduce the information loss caused by traditional FS and FR, short-term wind speed forecast with low information loss based on OSVD feature generation is proposed. First, the original wind speed series is denoised by OVMD. Then, based on the 96-dimensional original wind speed feature set, the OSVD is used to generate features. Furthermore, the extended original feature set EFS is obtained by combining the initial feature set with the features generated by OSVD. Gini importance is used to measure the importance of all features in EFS, and the forward feature selection is combined with random forests to determine the optimal subset. Finally, the optimal model determined by the new method is compared with seven models to verify the advancement of the new method. The experiments show that it reduces the information loss. Thus, the model has a higher forecast accuracy than the traditional model. INDEX TERMS Short-term wind speed forecast, feature selection, feature reduction, low information loss, variational mode decomposition, singular value decomposition. The associate editor coordinating the review of this manuscript and approving it for publication was Shuaihu Li. which makes the safe and reliable operation of a power grid enormously challenging and restricts the development of wind power. Accurate and efficient wind speed forecast (WSF) can reduce the negative impact of wind power uncertainty [4], [5]. WSF methods can generally be divided into physical methods [6], [7], statistical methods [8]-[11], artificial intelligence (AI) methods [12]-[17], and others. The physical method is based on a numerical weather prediction (NWP) model, which uses a series of meteorological data (wind direction, temperature, and humidity) and terrain information to establish WSFM [6], [7]. Unlike physical methods, statistical methods only use historical

Incorporating Load Fluctuation in Feature Importance Profile Clustering for Day-Ahead Aggregated Residential Load Forecasting

“…Because the RF is composed of multiple classification and regression trees (CARTs), it avoids unstable prediction results and overfitting. Additionally, the RF is a data-driven method based on ensemble learning theory; thus, it is effective for analyzing high-dimensional SM data [29].…”

“…To verify the improvements of the new method, the Pearson coefficient [38], mutual information (MI) index [39], and Gini index [29] are applied to the least-squares support vector machine (LS-SVM) [40], ANN [21], and extreme learning machine (ELM) [41] for experiments using the aforementioned modeling process. The clustering method is SDCKM.…”

“…Additionally, it screens the important features when DT generation occurs and its nodes split, reducing the effect of redundant features on the prediction accuracy [26]. Additionally, RFs have significant advantages to load forecasting with highdimensional feature sets [27]- [29].…”

Meta Learning-Based Hybrid Ensemble Approach for Short-Term Wind Speed Forecasting

“…According to the sampling time interval, WSF can be divided into short-term (timescales of minutes, hours), medium-term (timescales of days), and long-term prediction (timescales of months or years) [9]. In the literature, the proposed WSF methodologies include four categories, which are physical modeling methods, statistical methods, machine learning methods, and combined methods, respectively [10][11][12][13][14]. Physical modeling methods try to establish WSF model based on a series of meteorological data (such as wind direction, air pressure, temperature, and humidity et al) and topography information around wind farms (such as contour lines, roughness and obstacles) [15,16].…”

Combined Probability Prediction of Wind Power Considering the Conflict of Evaluation Indicators

“…A typical wind power probability prediction method usually constructs an optimal prediction model by selecting one predictor from a plurality of predictors based on data. The main methods are the nonparametric estimation method and the parameter estimation method [14], [15]. The parameter estimation method [15], [16] assumes that the prediction target obeys a certain distribution form, such as Gaussian distribution, warped Gaussian distribution, beta distribution, versatile distribution or logit-normal distribution.…”