An optimized LSTM model for clean coal ash content prediction in dense medium separation scenarios on the basis of the dual decomposition method DOI
Yongqi Liu,

Yuping Fan,

Xiaomin Ma

и другие.

International Journal of Coal Preparation and Utilization, Год журнала: 2024, Номер unknown, С. 1 - 26

Опубликована: Дек. 10, 2024

To increase the accuracy of clean coal ash content prediction during dense medium separation process and address time lag issue encountered when measuring content, a model based on WaOA-VMD-SGMD-WaOA-LSTM was proposed. The adopts dual decomposition techniques optimized Variational Mode Decomposition (VMD) Symplectic Geometric (SGMD), which can completely decompose original data, uses Walrus optimization algorithm (WaOA) to optimize hyperparameters Long Short-Term Memory (LSTM) model. In construction, characteristic data ore (𝑍2), raw (𝑍3), heavy mesoporous cyclone pressure (𝑍4), suspension density (𝑍5), magnetic (𝑍6) were combined with decomposed cleaned grouping S-IMF0~S-IMFn, CO-IMF1, CO-IMF2 as input variables construct multiple LSTM models. Finally, value is superimposed realize content. Based industrial preparation plant in Shanxi, China, results show that coefficient determination (R2) 0.9974. After adding secondary technology, average absolute error reduced by 60.99% compared single strategy.

Язык: Английский

A novel machine learning ensemble forecasting model based on mixed frequency technology and multi-objective optimization for carbon trading price DOI Creative Commons

Zejun Li,

Long Jun, Lue Li

и другие.

Frontiers in Energy Research, Год журнала: 2024, Номер 11

Опубликована: Янв. 11, 2024

Carbon trading prices are crucial for carbon emissions and transparent market pricing. Previous studies mainly focused on data mining in the prediction direction to quantify prices. Although prospect of high-frequency forecasting mechanisms is considerable, more mixed-frequency ensemble needed Therefore, this article designs a new type model increase scope research. The module divided into three parts: denoising, mixed frequency machine learning, multi-objective optimization, forecasting. Precisely, preprocessing technology enhanced by adopting self-attention mechanism can better remove noise extract effective features. Furthermore, introduced learning achieve comprehensive efficient prediction, evaluation criterion proposed measure optimal submodel. Finally, based deep strategy effectively integrate advantages low-frequency complex datasets. At same time, optimization algorithm optimize parameters model, significantly improving predictive ability integrated module. results four experiments Mean Absolute Percent Error index improved 28.3526% compared models, indicating that established address time distribution characteristics uncertainty issues predicted price which helps mitigate climate change develop low-carbon economy.

Язык: Английский

Процитировано

1
Data analysis and preprocessing techniques for air quality prediction: a survey DOI
Chengqing Yu, Jing Tan,

Yihan Cheng

и другие.

Stochastic Environmental Research and Risk Assessment, Год журнала: 2024, Номер 38(6), С. 2095 - 2117

Опубликована: Март 18, 2024

Язык: Английский

Процитировано

1
Volatility Interval Prediction of Crude Oil Spot Prices: An Improved Hybrid Model DOI
Jinliang Zhang, ziyi Liu

Опубликована: Янв. 1, 2024

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Язык: Английский

Процитировано

0
Volatility Interval Prediction of Crude Oil Spot Prices: An Improved Hybrid Model DOI
Jinliang Zhang, ziyi Liu

Опубликована: Янв. 1, 2024

Crude oil price volatility forecasting is important for energy policymaking and investment risk avoidance, which has attracted significant global attention. Due to the non-stationary non-linear characteristics of crude volatility, it a great challenge its forecasting. In order describe uncertainty provide richer information than point results, an improved hybrid interval prediction model containing signal processing, sequence complexity judgement, construction machine learning methods proposed. Firstly, original decomposed reconstructed into several new sequences with different by ICEEMDAN-FE. Secondly, are fuzzified obtain upper bounds lower interval. Then, high-frequency predicted IESN, low-frequency residual term ARIMA. Finally, The final result formed sum each bound result. WTI spot Brent selected analysis, consider influence futures on construction. effect intervals validated from pricing benchmark time scale dimensions, respectively. By analyzing reasons outstanding interval, can be judged that proposed provides idea prices.

Язык: Английский

Процитировано

0
Volatility Interval Prediction of Crude Oil Spot Prices: An Improved Hybrid Model DOI
Jinliang Zhang, ziyi Liu

Опубликована: Янв. 1, 2024

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Язык: Английский

Процитировано

0
A prediction model for chlorophyll concentration in seawater based on BorutaShap-MEMD-GRU DOI

Qiguang Zhu,

junjun yuan,

Linsong Zhang

и другие.

Physica Scripta, Год журнала: 2024, Номер 99(9), С. 096003 - 096003

Опубликована: Авг. 2, 2024

Abstract To solve the problem of difficulty in selecting multi-parameter features ocean and lack power traditional time-series prediction models predicting data, an GRU model based on Borutashap algorithm, a hybrid multivariate empirical modal decomposition is proposed to predict this paper. The feature selection multi-feature data carried out by algorithm XG-boost model, then selected are decomposed multi-modal decomposition, reconstructed get high-frequency low-frequency components, trend term components Permutation Entropy, finally respectively brought into network summed up final result. In paper, model’s effectiveness verified ablation experiments compared with other classical time series models, results show that has better effect.

Язык: Английский

Процитировано

0
A New Hybrid Velocity Prediction Model for Pipeline Detectors Based on Bo-Ssa, Bilstm, and the Attention Mechanism DOI
Junjie Ma, Yiming Li, Zhongchao Zhang

и другие.

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

0
PmForecast: leveraging temporal LSTM to deliver in situ air quality predictions DOI
Maryam Rahmani, Suzanne Crumeyrolle,

Nadége Allegri-Martiny

и другие.

Environmental Science and Pollution Research, Год журнала: 2024, Номер 31(39), С. 51760 - 51773

Опубликована: Авг. 10, 2024

Язык: Английский

Процитировано

0
Analysis of Global and Key PM2.5 Dynamic Mode Decomposition Based on the Koopman Method DOI Creative Commons

Yuhan Yu,

Dantong Liu, Bin Wang

и другие.

Atmosphere, Год журнала: 2024, Номер 15(9), С. 1091 - 1091

Опубликована: Сен. 8, 2024

Understanding the spatiotemporal dynamics of atmospheric PM2.5 concentration is highly challenging due to its evolution processes have complex and nonlinear patterns. Traditional mode decomposition methods struggle accurately capture features concentrations. In this study, we utilized global linearization capabilities Koopman method analyze hourly daily in Beijing–Tianjin–Hebei (BTH) region from 2019 2021. This approach decomposes data into superposition different spatial modes, revealing their hierarchical structure reconstructing dynamic processes. The results show that concentrations exhibit high-frequency cycles 12 24 h, as well low-frequency 124 353 days, while also modes growth, recession, oscillation. these enables reconstruction with a mean absolute percentage error (MAPE) only 0.6%. Unlike empirical (EMD), (KMD) avoids aliasing provides clearer identification key compared wavelet analysis. These findings underscore effectiveness KMD analyzing concentration, offering new insights understanding other phenomena.

Язык: Английский

Процитировано

0
A Novel Short-Term PM2.5 Forecasting Approach Using Secondary Decomposition and a Hybrid Deep Learning Model DOI Open Access

Ruru Liu,

XU Li-ping,

Tao Zeng

и другие.

Electronics, Год журнала: 2024, Номер 13(18), С. 3658 - 3658

Опубликована: Сен. 14, 2024

PM2.5 pollution poses an important threat to the atmospheric environment and human health. To precisely forecast concentration, this study presents innovative combined model: EMD-SE-GWO-VMD-ZCR-CNN-LSTM. First, empirical mode decomposition (EMD) is used decompose PM2.5, sample entropy (SE) assess subsequence complexity. Secondly, hyperparameters of variational (VMD) are optimized by Gray Wolf Optimization (GWO) algorithm, complex subsequences decomposed twice. Next, sequences divided into high-frequency low-frequency parts using zero crossing rate (ZCR); predicted a convolutional neural network (CNN), long short-term memory (LSTM). Finally, values reconstructed obtain final results. The experiment was conducted based on data 1009A, 1010A, 1011A from three air quality monitoring stations in Beijing area. results indicate that R2 value designed model increased 2.63%, 0.59%, 1.88% average stations, respectively, compared with other single mixed model, which verified significant advantages proposed model.

Язык: Английский

Процитировано

0