Short-term multi-hour ahead country-wide wind power prediction for Germany using gated recurrent unit deep learning

Renewable and Sustainable Energy Reviews, Journal Year: 2022, Volume and Issue: 167, P. 112700 - 112700

Published: June 24, 2022

Language: Английский

---

Short-term wind power prediction method based on CEEMDAN-GWO-Bi-LSTM

Energy Reports, Journal Year: 2024, Volume and Issue: 11, P. 1487 - 1502

Published: Jan. 18, 2024

In order to improve the short-term prediction accuracy of wind power and provide basis for grid dispatching, a complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) -grey wolf optimization (GWO) -bidirectional long memory network (Bi-LSTM) model is proposed predict output farms. Firstly, original data preprocessed, then decomposed into components that are easy extract features by using CEEMDAN. The Bi-LSTM established each component, grey algorithm used optimize parameters model. optimized hyperparameters brought results component. Finally, component superimposed reconstructed obtain final power. simulation analysis farm in Gansu Province shows CEEMDAN-GWO-Bi-LSTM has better prediction.

Language: Английский

---

Evaluating the Accuracy of the ERA5 Model in Predicting Wind Speeds Across Coastal and Offshore Regions

Journal of Marine Science and Engineering, Journal Year: 2025, Volume and Issue: 13(1), P. 149 - 149

Published: Jan. 16, 2025

Accurate wind speed and direction data are vital for coastal engineering, renewable energy, climate resilience, particularly in regions with sparse observational datasets. This study evaluates the ERA5 reanalysis model’s performance predicting speeds directions at ten offshore stations Kuwait from 2010 to 2017. analysis reveals that effectively captures general patterns, demonstrating stronger correlations (up 0.85) higher Perkins Skill Score (PSS) values 0.94). However, model consistently underestimates variability extreme events, especially stations, where correlation coefficients dropped 0.35. Wind highlighted ERA5’s ability replicate dominant northwest patterns. it notable biases underrepresented during transitional seasons. Taylor diagrams error metrics further emphasize challenges capturing localized dynamics influenced by land-sea interactions. Enhancements such as calibration using high-resolution datasets, hybrid models incorporating machine learning techniques, long-term monitoring networks recommended improve accuracy. By addressing these limitations, can more support engineering applications, including infrastructure design energy development, while advancing Kuwait’s sustainable development goals. provides valuable insights into refining complex environments.

Language: Английский

---

A novel decomposition-ensemble prediction model for ultra-short-term wind speed

Energy Conversion and Management, Journal Year: 2021, Volume and Issue: 248, P. 114775 - 114775

Published: Oct. 1, 2021

Language: Английский

---

Artificial Neural Networks Hidden Unit and Weight Connection Optimization by Quasi-Refection-Based Learning Artificial Bee Colony Algorithm

IEEE Access, Journal Year: 2021, Volume and Issue: 9, P. 169135 - 169155

Published: Jan. 1, 2021

Artificial neural networks are one of the most commonly used methods in machine learning. Performance network highly depends on learning method. Traditional algorithms prone to be trapped local optima and have slow convergence. At other hand, nature-inspired optimization proven very efficient complex problems solving due derivative-free solutions. Addressing issues traditional algorithms, this study, an enhanced version artificial bee colony metaheuristics is proposed optimize connection weights hidden units networks. Proposed improved method incorporates quasi-reflection-based guided best solution bounded mechanisms original approach manages conquer its deficiencies. First, tested a recent challenging CEC 2017 benchmark function set, then applied for training five well-known medical datasets. Further, devised algorithm compared metaheuristics-based methods. The efficiency measured by metrics - accuracy, specificity, sensitivity, geometric mean, area under curve. Simulation results prove that outperforms terms accuracy convergence speed. improvement over different datasets between 0.03% 12.94%. quasi-refection-based mechanism significantly improves speed together with bounded, exploitation capability enhanced, which better accuracy.

Language: Английский

---

A simple approach for short-term wind speed interval prediction based on independently recurrent neural networks and error probability distribution

Energy, Journal Year: 2021, Volume and Issue: 238, P. 122012 - 122012

Published: Sept. 8, 2021

Language: Английский

---

Energy forecasting model based on CNN-LSTM-AE for many time series with unequal lengths

Engineering Applications of Artificial Intelligence, Journal Year: 2022, Volume and Issue: 113, P. 104998 - 104998

Published: June 2, 2022

Language: Английский

---

Wind speed prediction using measurements from neighboring locations and combining the extreme learning machine and the AdaBoost algorithm

Energy Reports, Journal Year: 2022, Volume and Issue: 8, P. 1508 - 1518

Published: Jan. 6, 2022

Wind speed prediction plays an essential role in wind energy utilization. However, most existing studies of forecasting used data from one location to build models and forecasts, which limited the accuracy forecasting. Therefore, improve at a target location, this study proposes multiple-point model based on multiple locations for short-term prediction. The model, utilizes measurements neighboring combines extreme learning machine (ELM) with AdaBoost algorithm, is named multiple-point-AdaBoost-ELM model. Data seventeen automatic meteorological stations Heihe River Basin are used, four different positions taken as multi-time-scale prediction, six several metrics involved comparative analysis comprehensive evaluation. results show that: (1) performance proposed significantly superior that compared single-point models; (2) relatively less affected by time-scale than corresponding model; (3) located center can obtain more accurate those near edges region. promising method traditional modeling methods. fully uses historical surrounding enhance predictions locations, makes up deficiency using expands new way

Language: Английский

---

A novel approach to ultra-short-term wind power prediction based on feature engineering and informer

Energy Reports, Journal Year: 2022, Volume and Issue: 9, P. 1236 - 1250

Published: Dec. 26, 2022

Wind power is prone to dramatic fluctuations in the short term, posing a threat safety and stability of grid, so accurate forecasting ultra-short-term wind important ensure economy system. The historical data an enormous nonlinear time series. It expected mine independent features related from original through feature engineering, then use Informer model solve prediction problem long-time series power, thus reducing space complexity improving accuracy. In this paper, Turkey farm are selected predict 10 min based on engineering model. First, factor with high correlation formed after engineering. Then, train conduct multiple experiments obtain optimal parameters. Finally, results compared recurrent neural network (RNN) model, long-short-term memory (LSTM) Transformer experimental show that has accuracy operation efficiency. Four evaluation metrics mean absolute error, square symmetric percentage runtime decreased by at least 32.849, 8495.193, 5.544%, 92, which proves approach prominent advantages prediction. Its can provide reference for coordinated dispatching, risk analysis, scientific decision-making systems.

Language: Английский

---

A new hybrid optimization prediction model for PM2.5 concentration considering other air pollutants and meteorological conditions

Chemosphere, Journal Year: 2022, Volume and Issue: 307, P. 135798 - 135798

Published: Aug. 11, 2022

Language: Английский

---