A Temporal Network Based on Characterizing and Extracting Time Series in Copper Smelting for Predicting Matte Grade DOI Creative Commons

Junjia Zhang,

Zhuorui Li, Enzhi Wang

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(23), P. 7492 - 7492

Published: Nov. 24, 2024

Addressing the issues of low prediction accuracy and poor interpretability in traditional matte grade models, which rely on pre-smelting input assay data for regression, we incorporate process sensors' propose a temporal network based Time to Vector (Time2Vec) convolutional combined with multi-head attention (TCN-TMHA) tackle weak characteristics uncertain periodic information copper smelting process. Firstly, employed maximum coefficient (MIC) criterion select strongly correlated grade. Secondly, used Time2Vec module extract from variables, incorporates time series processing directly into model. Finally, implemented TCN-TMHA specific weighting mechanisms assign weights features prioritize relevant key step features. Experimental results indicate that proposed model yields more accurate predictions content, determination (

Language: Английский

Tight Oil Well Productivity Prediction Model Based on Neural Network DOI Open Access

Yuhang Jin,

Kangliang Guo,

Xinchen Gao

et al.

Processes, Journal Year: 2024, Volume and Issue: 12(10), P. 2088 - 2088

Published: Sept. 26, 2024

Productivity prediction has always been an important part of reservoir development, and tight reservoirs need accurate efficient productivity models. Due to the complexity oil reservoir, data obtained by detection instrument extract features at a deeper level. Using Pearson correlation coefficient partial analyze main control factors, eight characteristic parameters volume coefficient, water saturation, density, effective thickness, skin factor, shale content, porosity, permeability were obtained, specific production index was used as target parameter. Two sample structures pure static dynamic (shale permeability, density parameters, thickness parameters) created, corresponding model (BP (Backpropagation), neural network model, LSTM-BP (Long Short-Term Memory Backpropagation) model) designed compare effects models under different structures. The mean absolute error, root square relative percentage determination evaluate results. predict capacity test set. results showed that average error 0.07, 0.10, 21%, 0.97. wells in WZ area for testing, model’s predictions are evenly distributed on both sides 45° line, separating predicted values from actual values, with errors line being relatively small. In contrast, BP analytical method unable achieve such even distribution around line. Experiments show can effectively parameter stronger generalization ability.

Language: Английский

Citations

3

A Hybrid Tabular-Spatial-Temporal Model with 3D Geo-Model for Production Prediction in Shale Gas Formations DOI
Muming Wang, Hai Wang, Shengnan Chen

et al.

SPE Annual Technical Conference and Exhibition, Journal Year: 2024, Volume and Issue: 191

Published: Sept. 20, 2024

Abstract The evolution of shale gas production has reshaped North America's energy profile. Utilizing the vast amounts data generated from and operations, machine learning offers significant advantages in forecasting performance optimization. This study proposed a pioneering hybrid model integrating tabular, spatial, temporal modalities to enhance unconventional reservoirs. Despite traditional methods such as artificial neural networks (ANN) XGBoost, which rely solely on tabular for training prediction, this proposes novel 3D-parameterization method. approach tokenizes formation property distribution into 3-axis tensors, enabling more comprehensive representation spatial data. Then, 3D-convolutional network (3D-CNN) with attention mechanism module was established process created For modality, long short-term memory (LSTM) used accept dynamic input predict monthly simultaneously. A total 677 wells Duvernay collected, pre-processed fed according based their modality. results show that combined three achieved an impressive level accuracy, coefficient determination (R2) 0.8771, surpassing (0.7841) tabular-spatial (0.8230) models. Additionally, global optimization applied further by optimizing architecture each hyperparameters, 1.88% improvement empirical design. These advancements set new benchmark predictive modelling reservoirs, highlighting importance utilizing different improving forecast prediction.

Language: Английский

Citations

1

A flow rate estimation method for gas–liquid two-phase flow based on filter-enhanced convolutional neural network DOI
Yuxiao Jiang, Yinyan Liu, Lihui Peng

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 139, P. 109593 - 109593

Published: Nov. 11, 2024

Language: Английский

Citations

1

PMTT: Parallel multi-scale temporal convolution network and transformer for predicting the time to aging failure of software systems DOI
Kai Jia, Xiao Yu, Chen Zhang

et al.

Journal of Systems and Software, Journal Year: 2024, Volume and Issue: 217, P. 112167 - 112167

Published: July 31, 2024

Language: Английский

Citations

0

A Hybrid Prediction Model for International Crude Oil Price Based on Variational Mode Decomposition with BiTCN-BiGRU-Attention Deep Learning Techniques DOI Creative Commons
Meihua Bi, Ziyun Liu, Xiaozhong Yang

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 21, 2024

Abstract Predicting the price and volatility of international crude oil futures is a complex task. This paper presents novel hybrid prediction model, VMD-BiTCN-BiGRU-Attention, which integrates variational mode decomposition (VMD) advanced deep learning techniques to forecast nonlinear, non-stationary, time-varying characteristics sequences. Initially, sequence decomposed into multiple modes using VMD, enabling capture different frequency components. Each independently predicted bidirectional time convolutional network (BiTCN), captures temporal information enhances long-term dependencies through dilated convolution. Subsequently, gated recurrent unit (BiGRU) models more effectively, while an attention mechanism adjusts weights BiGRU outputs emphasize critical information. The model’s predictions are optimized with Adam algorithm. Empirical results demonstrate that model adept at forecasting non-stationary nonlinear prices. Furthermore, Diebold-Mariano (DM) test confirms this surpasses 15 other regarding accuracy performance, achieving optimal key metrics: R² = 0.9953, RMSE 1.4417, MAE 0.7973, MAPE 1.5213%. These findings underscore its potential for enhancing prediction.

Language: Английский

Citations

0

Single Well Production Prediction Model of Gas Reservoir Based on CNN-BILSTM-AM DOI Creative Commons

Daihong Gu,

Rongchen Zheng, Peng Cheng

et al.

Energies, Journal Year: 2024, Volume and Issue: 17(22), P. 5674 - 5674

Published: Nov. 13, 2024

In the prediction of single-well production in gas reservoirs, traditional empirical formula reservoirs generally shows poor accuracy. process machine learning training and prediction, problems small data volume dirty are often encountered. order to overcome above problems, a model based on CNN-BILSTM-AM is proposed. The built by long-term short-term memory neural networks, convolutional networks attention modules. input includes previous period its influencing factors. At same time, fitting error value reservoir introduced predict future data. loss function used evaluate deviation between predicted real data, Bayesian hyperparameter optimization algorithm optimize structure comprehensively improve generalization ability model. Three single wells Daniudi D28 well area were selected as database, was production. results show that compared with network (CNN) model, long (LSTM) bidirectional (BILSTM) test set three experimental reduced 6.2425%, 4.9522% 3.0750% average. It basis coupling meets high-precision requirements for which great significance guide efficient development oil fields ensure safety China’s energy strategy.

Language: Английский

Citations

0

A Temporal Network Based on Characterizing and Extracting Time Series in Copper Smelting for Predicting Matte Grade DOI Creative Commons

Junjia Zhang,

Zhuorui Li, Enzhi Wang

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(23), P. 7492 - 7492

Published: Nov. 24, 2024

Addressing the issues of low prediction accuracy and poor interpretability in traditional matte grade models, which rely on pre-smelting input assay data for regression, we incorporate process sensors' propose a temporal network based Time to Vector (Time2Vec) convolutional combined with multi-head attention (TCN-TMHA) tackle weak characteristics uncertain periodic information copper smelting process. Firstly, employed maximum coefficient (MIC) criterion select strongly correlated grade. Secondly, used Time2Vec module extract from variables, incorporates time series processing directly into model. Finally, implemented TCN-TMHA specific weighting mechanisms assign weights features prioritize relevant key step features. Experimental results indicate that proposed model yields more accurate predictions content, determination (

Language: Английский

Citations

0