Deep Learning Framework for Accurate Static and Dynamic Prediction of CO2 Enhanced Oil Recovery and Storage Capacity DOI Open Access

Zhipeng Xiao,

Bin Shen,

Jiguang Yang

et al.

Processes, Journal Year: 2024, Volume and Issue: 12(8), P. 1693 - 1693

Published: Aug. 13, 2024

As global warming intensifies, carbon capture, utilization, and storage (CCUS) technology is widely used to reduce greenhouse gas emissions. CO2-enhanced oil recovery (CO2-EOR) has, once again, received attention, which can achieve the dual benefits of CO2 storage. However, flexibly effectively predicting flooding capacity potential reservoirs a major problem. Traditional prediction methods often lack ability comprehensively integrate static dynamic predictions and, thus, cannot fully understand CO2-EOR capacity. This study proposes comprehensive deep learning framework, named LightTrans, based on lightweight gradient boosting machine (LightGBM) Temporal Fusion Transformers, for The model predicts cumulative production, amount, Net Present Value test set with an average R-square (R2) 0.9482 mean absolute percentage error (MAPE) 0.0143. It shows great performance. In addition, its R2 0.9998, MAPE 0.0025. excellent ability. proposed successfully captures time-varying characteristics systems. worth noting that our 105–106 times faster than traditional numerical simulators, again demonstrates high-efficiency value LightTrans model. Our framework provides efficient, reliable, intelligent solution development optimization

Language: Английский

Comprehensive Assessment of E. coli Dynamics in River Water Using Advanced Machine Learning and Explainable AI DOI

Santanu Mallik,

Bikram Saha,

Krishanu Podder

et al.

Process Safety and Environmental Protection, Journal Year: 2025, Volume and Issue: unknown, P. 106816 - 106816

Published: Jan. 1, 2025

Language: Английский

Citations

1
Boosting-Based Machine Learning Applications in Polymer Science: A Review DOI Open Access
Ivan Malashin, В С Тынченко, Andrei Gantimurov

et al.

Polymers, Journal Year: 2025, Volume and Issue: 17(4), P. 499 - 499

Published: Feb. 14, 2025

The increasing complexity of polymer systems in both experimental and computational studies has led to an expanding interest machine learning (ML) methods aid data analysis, material design, predictive modeling. Among the various ML approaches, boosting methods, including AdaBoost, Gradient Boosting, XGBoost, CatBoost LightGBM, have emerged as powerful tools for tackling high-dimensional complex problems science. This paper provides overview applications science, highlighting their contributions areas such structure-property relationships, synthesis, performance prediction, characterization. By examining recent case on techniques this review aims highlight potential advancing characterization, optimization materials.

Language: Английский

Citations

1
Edge-cloud collaboration-driven predictive planning based on LSTM-attention for wastewater treatment DOI
Shuaiyin Ma, Wei Ding,

Yujuan Zheng

et al.

Computers & Industrial Engineering, Journal Year: 2024, Volume and Issue: 195, P. 110425 - 110425

Published: July 27, 2024

Language: Английский

Citations

6
Multidimensional Lost Circulation Risk Quantification Assessment Model Based on Ensemble Machine Learning DOI
Haibo Mu, Guancheng Jiang, Wei Zhang

et al.

SPE Journal, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 11

Published: March 1, 2025

Summary The risk of lost circulation is a complex problem that cannot be ignored during drilling operations, and accurate assessment crucial for preventing controlling events. In this study, we establish multidimensional quantitative model based on ensemble machine learning, comprehensively considering three dimensions—formation risk, operation fluid risk. It can effectively capture quantify the interactive relationship between different factors, accuracy efficiency improved when learning algorithms determine dimensional weights. results example verification show threshold index set to 0.55, in 442 samples drilled certain block, 85.02% without 70.21% with circulation. This result reflects uncertainty occurrence events field difference two categories approximately 15%, error within an acceptable range (0.1~0.2). independent variable parameters each dimension adjusted according actual situation blocks, thresholds correction factors set. established has high adaptability, which guide prevention control.

Language: Английский

Citations

0
Physics-Informed Dynamic Bayesian Networks for Time-Dependent Reliability Prediction of Subsea Wellhead Sealing System with Multi-States DOI
Shengnan Wu, Huarong Gong, Shengtao Ma

et al.

Published: Jan. 1, 2025

Language: Английский

Citations

0
Analyzing Key Parameters in Underground Hydrogen Storage Using Machine Learning Surrogate Models DOI

Tanin Esfandi,

Yasin Noruzi,

Mir Saeid Safavi

et al.

Published: Jan. 1, 2025

Language: Английский

Citations

0
Advancement of artificial intelligence applications in hydrocarbon well drilling technology: A review DOI
Shadfar Davoodi, Mohammed Al-Shargabi, David A. Wood

et al.

Applied Soft Computing, Journal Year: 2025, Volume and Issue: unknown, P. 113129 - 113129

Published: April 1, 2025

Language: Английский

Citations

0
Deep Learning Framework for Accurate Static and Dynamic Prediction of CO2 Enhanced Oil Recovery and Storage Capacity DOI Open Access

Zhipeng Xiao,

Bin Shen,

Jiguang Yang

et al.

Processes, Journal Year: 2024, Volume and Issue: 12(8), P. 1693 - 1693

Published: Aug. 13, 2024

As global warming intensifies, carbon capture, utilization, and storage (CCUS) technology is widely used to reduce greenhouse gas emissions. CO2-enhanced oil recovery (CO2-EOR) has, once again, received attention, which can achieve the dual benefits of CO2 storage. However, flexibly effectively predicting flooding capacity potential reservoirs a major problem. Traditional prediction methods often lack ability comprehensively integrate static dynamic predictions and, thus, cannot fully understand CO2-EOR capacity. This study proposes comprehensive deep learning framework, named LightTrans, based on lightweight gradient boosting machine (LightGBM) Temporal Fusion Transformers, for The model predicts cumulative production, amount, Net Present Value test set with an average R-square (R2) 0.9482 mean absolute percentage error (MAPE) 0.0143. It shows great performance. In addition, its R2 0.9998, MAPE 0.0025. excellent ability. proposed successfully captures time-varying characteristics systems. worth noting that our 105–106 times faster than traditional numerical simulators, again demonstrates high-efficiency value LightTrans model. Our framework provides efficient, reliable, intelligent solution development optimization

Language: Английский

Citations

2