A Wind Speed Prediction Method Based on Signal Decomposition Technology Deep Learning Model

Energies, Год журнала: 2025, Номер 18(5), С. 1136 - 1136

Опубликована: Фев. 25, 2025

Accurate and reliable wind speed prediction plays a significant role in ensuring the reasonable scheduling of power resources. However, sequences often exhibit complex characteristics such as instability volatility, which create substantial challenges for prediction. In order to cope with these challenges, multi-step method based on secondary decomposition (SD) techniques deep learning models is proposed this paper. First, original signal was decomposed into multiple by using two techniques, multi-scale wavelet spectrum analysis (MWPSA) variational mode (VMD). Second, model constructed combining convolutional neural networks (CNNs), bidirectional long short-term memory (BiLSTM) networks, attention mechanism perform predicting each sequence, parameters were optimized particle swarm optimization (PSO) algorithm. Ultimately, results from all combined generate final The predictive performance evaluated real data collected farm China. Experimental show that significantly outperforms other comparison prediction, highlights its accuracy reliability.

Язык: Английский

---

An improved GRU method for slope stress prediction

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 21, 2025

The stability of open-pit mine slopes is a complex nonlinear system. Stress variation significant influencing factor in the occurrence landslide disasters and also key research focus early warning risk assessment. However, traditional methods are confronted with challenges, including low prediction accuracy poor robustness when dealing time series data. In order to address aforementioned issues, present paper proposes an intelligent model based on Variational Mode Decomposition (VMD) Dung Beetle Optimization (DBO), combined improved Gated Recurrent Unit (GRU), which hereby referred as VMD-DBO-GRU-A model. preliminary preprocessing open pit slope stress data using VMD can provide high decomposition effectively extract localized features stress; method introduces (DBO) determine number hidden neuron layers optimal learning rate for GRU. This reduces uncertainty parameters minimizes required parameter tuning; Self-attention mechanism added assign different weights input features, dependence external information more adept at capturing internal relevance or features. verify validity model, experiments conducted self-constructed dataset this paper. experimental results show that root-mean-square error has decreased by 77% 84% compared LSTM SVM models, respectively, coefficient determination 0.9978, fully verifies excellent comprehensive performance accuracy, great value practical application mines' slopes.

Язык: Английский

---

An embedded spatiotemporal hybrid model integrating multi-graphs and attention-driven fusion for single- and multi-site photovoltaic power forecasting

Energy Conversion and Management, Год журнала: 2025, Номер 336, С. 119897 - 119897

Опубликована: Май 8, 2025

Язык: Английский

---

An interpretable interval-valued wind power prediction system based on multi-objective feature extraction and base model selection with dynamic ensemble

Swarm and Evolutionary Computation, Год журнала: 2025, Номер 96, С. 101977 - 101977

Опубликована: Май 9, 2025

Язык: Английский

---

An integrated prediction model based on meta ensemble learning for short‐term wind speed forecasting

IET Renewable Power Generation, Год журнала: 2024, Номер unknown

Опубликована: Июнь 28, 2024

Abstract As wind power increasingly integrates into grids and energy systems, accurate reliable speed forecasting (WSF) has become essential for scheduling management. Considering the fluctuating random characteristics of speed, a novel integrated model short‐term WSF is developed in this work, which multiple models through meta ensemble learning to achieve better generalization, robustness, accuracy. This consists four components: data input, base predictor, learning, prediction output. The predictor component includes pre‐trained predictors long memory recurrent neural network provide initial values speed. multi‐input multi‐output back propagation that outputs automatically adjust weight coefficients various predictors, drawing on environmental meteorological historical data. ultimate result obtained weighted summation predictors. authors assess effectiveness by contrasting its performance with alternative models. simulation results reveal proposed surpasses both individual traditional approaches terms stability accuracy WSF.

Язык: Английский

---

Data denoising and deep learning prediction for the wind speed based on NOA optimization

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Авг. 2, 2024

Accurate short-term wind speed prediction is of great significance for power generation. Due to the insufficient traditional methods mine nonlinear features information, an improved time series method proposed by combining Variational Mode Decomposition (VMD) and Deep Learning (CNN-BiLSTM-AttNTS) with Nutcracker Optimization Algorithm (NOA). Firstly, NOA used optimize VMD CNN-BiLSTM, respectively. Secondly, we apply NOA-VMD decompose data into different Intrinsic Functions(IMFs). Then, phase space reconstruction (PSR) utilized identify chaotic characteristics components. Finally, NOA-CNN-BiLSTM-AttNTS model built up predict speed. Under same hyperparameters network structure settings, compared machine learning state-of-the-art hybrid models, results show that R-squared NOA-VMD-CNN-BiLSTM-AttNTS combination in this paper exceeds 90%, good accuracy generalization performance. The research result can provide reference guidance prediction.

Язык: Английский

---

A singular spectrum analysis-enhanced BiTCN-selfattention model for runoff prediction

Earth Science Informatics, Год журнала: 2024, Номер 18(1)

Опубликована: Дек. 12, 2024

Язык: Английский

---

A comprehensive framework of the decomposition-based hybrid method for ultra-short-term wind power forecasting with on-site application

Energy, Год журнала: 2024, Номер unknown, С. 133911 - 133911

Опубликована: Ноя. 1, 2024

Язык: Английский

---

A novel analysis of random forest regression model for wind speed forecasting

Cogent Engineering, Год журнала: 2024, Номер 11(1)

Опубликована: Дек. 6, 2024

This article uses a random forest regression (RFR) model to analyze wind speed forecasting. Wind energy is one of the more critical potentials in renewable sources for producing clean and safe environment. Accurate stable forecasting essential improving efficiency turbines, guaranteeing power balance, economic dispatch systems ensuring equipment safety. Previous researchers have attempted address these issues less prediction performance lack interpretable analysis. study aims develop machine learning (ML) models, such as neural networks (NNs), linear (LR), support vector (SVR), decision tree (DTR), K-nearest neighbors (K-NN), extreme gradient boosting RFR. Six evaluation criteria are applied estimate ML model: mean squared error, root absolute error (MAE), percentage normalized average squares coefficient determination. The experimental results show RFR achieves better accuracy than other models. from was NMSE = 0.003, MAE 0.049, MSE 0.033, RMSE 0.182, MAPE 1.180 R2 0.996. Precise predictions various industries, aviation, shipping generation.

Язык: Английский

---

An innovative memory-enhanced Elman neural network-based selective ensemble system for short-term wind speed prediction

Applied Energy, Год журнала: 2024, Номер 380, С. 125108 - 125108

Опубликована: Дек. 12, 2024

Язык: Английский

---