Self‐paced learning long short‐term memory based on intelligent optimization for robust wind power prediction

IET Control Theory and Applications, Год журнала: 2024, Номер unknown

Опубликована: Апрель 3, 2024

Abstract Given the unpredictable and intermittent nature of wind energy, precise forecasting power is crucial for ensuring safe stable operation systems. To reduce influence noise data on robustness prediction, a prediction method proposed that leverages an enhanced multi‐objective sand cat swarm algorithm (MO‐SCSO) self‐paced long short‐term memory network (spLSTM). First, actual processed into time series as input output. Then, progressive advantage learning used to effectively solve instability caused by noisy during (LSTM) training. Following this, improved MO‐SCSO employed iteratively optimize hyperparameters spLSTM. Ultimately, combined MO‐SCSO‐spLSTM model constructed prediction. This validated with onshore farms in Austria offshore Denmark. The experimental results show compared traditional LSTM method, has better accuracy robustness. Specifically, experiments, reduces minimum MAE 5.44% 4.96%, respectively, range 4.45% 17.21%, which could be conducive system.

Язык: Английский

---

MSVMD-Informer: A Multi-Variate Multi-Scale Method to Wind Power Prediction

Energies, Год журнала: 2025, Номер 18(7), С. 1571 - 1571

Опубликована: Март 21, 2025

Wind power prediction plays a crucial role in enhancing grid stability and wind energy utilization efficiency. Existing methods demonstrate insufficient integration of multi-variate features, such as speed, temperature, humidity, along with inadequate extraction correlations between variables. This paper proposes novel multi-scale method named variational mode decomposition informer (MSVMD-Informer). First, modal module is designed to decompose univariate time-series features into multiple scales. Adaptive graph convolution applied extract scales, while self-attention mechanisms are utilized capture temporal dependencies within the same scale. Subsequently, feature fusion proposed better account for inter-variable correlations. Finally, reconstructed by integrating aforementioned modules, enabling forecasting. The was evaluated through comparative experiments ablation studies against seven baselines using public dataset two private datasets. Experimental results that our achieves optimal metric performance, its lowest MAPE scores being 1.325%, 1.500% 1.450%, respectively.

Язык: Английский

---

Systematic Performance Analysis of Long Short-Term Memory Neural Network for Wind Speed Predictions

Опубликована: Янв. 1, 2025

Язык: Английский

---

Fatigue Load Modeling of Floating Wind Turbines Based on Vine Copula Theory and Machine Learning

Journal of Marine Science and Engineering, Год журнала: 2024, Номер 12(8), С. 1275 - 1275

Опубликована: Июль 29, 2024

Fatigue load modeling is crucial for optimizing and assessing the lifespan of floating wind turbines. This study addresses complex characteristics fatigue loads on turbines under combined effects waves. We propose a approach based Vine copula theory machine learning. Firstly, we establish an optimal joint probability distribution model using four-dimensional random variables (wind speed, wave height, period, direction), with fit assessed Akaike Information Criterion (AIC), Bayesian (BIC), Root Mean Square Error (RMSE). Secondly, representative conditions are determined Monte Carlo sampling established model. Thirdly, simulations performed high-fidelity simulator OpenFAST to compute Damage Equivalent Load (DEL) values critical components (blade root tower base). Finally, utilizing measured data from Lianyungang Ocean Observatory in East China Sea, simulation tests conducted. apply five commonly used learning models (Kriging, MLP, SVR, BNN, RF) develop DEL blade base. The results indicate that RF exhibits smallest prediction error, not exceeding 3.9%, demonstrates high accuracy, particularly predicting flapwise at root, achieving accuracies up 99.97%. These findings underscore effectiveness our accurately real-world conditions, which essential enhancing reliability efficiency

Язык: Английский

---

Wind Power Ramp Events Prediction Considering Wind Propagation

Renewable Energy, Год журнала: 2024, Номер 236, С. 121280 - 121280

Опубликована: Сен. 6, 2024

Язык: Английский

---

Optimizing Wind Power Forecasting with RNN-LSTM Models through Grid Search Cross-validation

Sustainable Computing Informatics and Systems, Год журнала: 2024, Номер unknown, С. 101054 - 101054

Опубликована: Ноя. 1, 2024

Язык: Английский

---