Current status of hybrid structures in wind forecasting. (March 2021)

• Record Type: Journal Article
• Authors: Ahmadi, Mehrnaz; Khashei, Mehdi
• Is Part Of: Engineering applications of artificial intelligence. Volume 99(2021)
• Publication Date: 2021-03
• Subjects: AAE Adversarial Auto-Encoder -- ACO Ant Colony Optimization -- ADALINE Adaptive Linear Element Network -- AELM AdaBoost ELM -- AnEn A novelty the application of an analog Ensemble -- ARIMA Autoregressive Integrated Moving Average -- ASD Adaptive Secondary Decomposition -- ANN Artificial Neural Network -- AWNN Adaptive Wavelet Neural Network -- ANFIS Adaptive Neuro Fuzzy Inference System -- BFGS Broyden–Fletcher–Goldfarb–Shanno -- BCF A combined algorithm based on BA, CS and FA -- BSA Backtracking Search Algorithm -- BMA Bayesian Model Averaging -- BP Back Propagation neural network -- CBP Cascade BP neural network -- CEEMDAN Complete Ensemble Empirical Mode Decomposition Adaptive Noise -- CNN Convolutional Neural Network -- DSS Double Similarity Search -- CS Cuckoo Search algorithm -- CLSFPA Chaotic Local Search -- CSA Coupled Simulated Annealing -- CVR Core Vector Regression model -- COR Competition Over Resource -- CG-BP-FR Conjugate Gradient Back Propagation with Fletcher–Reeves updates -- CNNSVM Convolutional Support Vector Machine -- CT Correction Topography -- COSMO Consortium for Small-scale Modeling -- CQR Composite Quantile Regression -- CH-MKL Heteroscedastic Multi-Kernel Learning based Correlation Aware Multi-output Regression -- COSMO-LEPS Limited-area Ensemble Prediction System developed within Consortium for Small-scale Modeling -- DRL Deep Reinforcement learning -- DBN Deep Belief Network model -- DBSCAN Density-Based Spatial Clustering of Applications with Noise -- DWT Discrete Wavelet Transform -- EMD Empirical Mode Decomposition -- EFG Enhanced Forget-Gate network model -- ENN Elman Neural Network -- ECMWF European Centre for Medium-Range Weather Forecasts Ensemble Prediction System -- EWT Empirical Wavelet Transform -- ECS Error Correction Strategy -- ESN Echo State Network -- ELM Extreme Learning Machine -- GARCH Generalized Autoregressive Conditionally Heteroscedastic -- FAC First-order Adaptive Coefficient -- FPA Flower-Pollination Algorithm -- FEEMD Fast Ensemble Empirical Mode Decomposition -- FN Functional Networks -- FOMC First Order Markov Chain -- GRU Gated Recurrent Units network -- GWO The Gray Wolf Optimizer algorithm -- GD-ALR-BP Gradient Descent with Adaptive Learning Rate Back Propagation -- GSA Gravitational Search Algorithm -- GDM-ALR-BP Gradient Descent with Momentum and Adaptive Learning Rate Back Propagation -- GPR Gaussian process Regression method -- GBR Gradient Boosted Regression -- XGB Extreme Gradient Boosting -- GSO Gram–Schmidt Orthogonal -- GWPPT Generalized Wind Power Prediction Tool -- HELM Hysteretic Extreme Learning Machine -- HAR Hammerstein Auto-Regressive model -- HBSA Hybrid Backtracking Search Algorithm -- HMD Hybrid Mode Decomposition method -- IEWT Inverse Empirical Wavelet Transform -- WPD Wavelet Packet Decomposition -- ICA Imperialist Competitive Algorithm -- ISMC Indexed Semi-Markov Chain model -- IBP Improved Back-Propagation neural network -- ICEEMDAN Improved Complementary Ensemble Empirical Mode Decomposition with Adaptive Noise -- KDE novel Kernel Density Estimator -- KPCA Kernel Principal Component Analysis -- KFCM kernel-based Fuzzy C-Means clustering algorithm -- LLNF Local Linear Fuzzy Neural network -- LRELM a Leave-one-out cross-validation-based Regularized Extreme Learning Machine -- LSTM Long Short-Term Memory -- RELM Regularized Extreme Learning Machine network -- RVM Relevance Vector Machines -- LS-SVM Least Squares-Support Vector Machine -- MKRPINN Multi-Kernel Regularized Pseudo Inverse Neural Network -- MAdaBoost Modified AdaBoost -- MHF Morphological High-Frequency filter -- MOSCA Multi-Objective Sine Cosine Algorithm -- MLFFNN Multilayer Feed-Forward Neural Network -- MOALO Multi-Objective Ant Lion Algorithm -- MWDO Modified Wind Driven Optimization model -- MSS Model Structure Selection -- MKRR Multi-Kernel Ridge Regression -- MADF Manifold Algorithms used for Data Fusion method -- MWLR Modified WLR -- NARX Nonlinear Autoregressive with Exogenous inputs network -- NNCT No Negative Constraint Theory -- NCL-RELM Negative Correlation Learning-based Regularized Extreme Learning Machine -- NCFM Three hybrid models (EMD-BA-BPNN, EMD-BA-ENN and EMD-ARIMA) -- NNS Nearest Neighbor Search -- N-SVR Technique of the uniform model of ν-Support Vector Regression for the general noise model -- OSORELM Online Sequential Outlier Robust Extreme Learning Machine -- OFE Optimal Feature Extraction -- OVMD Optimized Variational Mode Decomposition -- ORELM Outlier-Robust Extreme Learning Machine -- OMWGP Online Model selection and the Warped Gaussian Process -- PSO Particle Swarm Optimization -- PACE Partial Auto-Correlation function -- PNN Polynomial Neural Networks -- PSOGSA Partial Swarm Optimization combined with Gravitational Search Algorithm -- PLS Partial Least Square -- PSR Phase Space Reconstruction -- QPSO Quantum behaved Particle Swarm Optimization -- QRNNE-UCV Quantile Regression Neural Network and Epanechnikov kernel function using Unbiased Cross-Validation -- RSVM Reduced Support Vector Machine -- RFR Random Forest Regression -- SR-NN Sinusoidal Rough-Neural Network -- SSA Singular Spectrum Analysis -- SE Sample Entropy -- SVR-RBF Support Vector Regression with a Radial Basis Function -- SR-FR Sparse Bayesian-based robust Functional Regression -- SWGP Sparse online Warped Gaussian Process -- SAM Hybrids the Seasonal Adjustment Method -- SPSA Simultaneous Perturbation Stochastic Approximation -- SOMC Second Order Markov Chain -- T2FNN Type-2 Fuzzy Neural Network -- TDRF Top-Down Relevant Feature search algorithm -- TW-FE Time-vary-Forecasting-Effectiveness -- TVARX Time Varying threshold Autoregressive model with interactions -- TARCHX Threshold seasonal Autoregressive Conditional Heteroscedastic model -- TLBO Teaching Learning Based Optimization -- VMD Variational Mode Decomposition -- VAR Vector Auto-Regression -- VAR-mGARCH Vector Autoregressive with multivariate Autoregressive Conditional Heteroskedasticity -- SampEn Sample Entropy -- WF Wavelet Filter -- WCA Water Cycle Algorithm -- WDD Wavelet Domain Denoising; Wind power forecasting -- Wind speed forecasting -- Hybrid forecasting approaches -- Series hybrid models -- Parallel hybrid models -- Time series forecasting; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2020.104133
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Wind power is one of the most important clean energy and alternative to fossil fuels. More attention has been paid to this renewable resource in today's world due to increasing public awareness, concerns about greenhouse gas emissions and environmental issues, and reducing the oil and gas reservoirs. Accurate and precise wind speed and wind power forecasts are the most critical and influential factors in making desired and efficient managerial and operational decisions in the wind energy area. Wind power and speed forecasting play an essential role in the planning, controlling, and monitoring of intelligent wind power systems. Therefore, several different models have been developed in the subject literature in order to predict this energy source more accurately. However, there is no general consensus on the model that must be selected and used in a specific situation of time horizon, sample size, complexity, uncertainty, etc. Hybrid models are the most frequently used and the most popular forecasting models in the energy literature. In this paper, combined approaches used in the wind energy forecasting field are first categorized into four main categories: 1) Data preprocessing based approaches, 2) Parameter optimization-based approaches, 3) Post processing based approaches, and 4) component combination-based approaches. Results indicate that the component combination-based category is the most diverse and extensive hybrid approach in the literature. Thus, in the next section of the paper, more attention is paid to these approaches and then classified into two major classes of series and parallel hybrid models. The literature review demonstrates that parallel hybrid models are more popular approaches in comparison with series hybrid models and more used for wind forecasting. Other specific and detailed conclusions and remarks are introduced in related sections.