Article
Thermodynamics
Shuyi Du, Meizhu Wang, Jiaosheng Yang, Yang Zhao, Jiulong Wang, Ming Yue, Chiyu Xie, Hongqing Song
Summary: This study proposes an enhanced prediction framework for coalbed methane (CBM) production forecasting using bidirectional long short-term memory (BiLSTM) and transfer learning. The findings demonstrate that Bi-LSTM has stronger memory capability and can improve the prediction accuracy of CBM dynamic data. The application of transfer learning significantly enhances the prediction performance for horizontal wells.
Article
Engineering, Environmental
Chao Li, Yong Qin, Tao Guo, Jian Shen, Yi Yang
Summary: This study measured the high-pressure adsorption isotherms of two coal samples to investigate the methane adsorption process and mechanisms in coal. The results showed that micropore filling and surface adsorption are the primary adsorption mechanisms for supercritical methane in coal, with micropore filling occurring in pores smaller than 1.1 nm and surface adsorption occurring in 1-2 nm pores. Micropore filling adsorption plays a significant role in the overall methane adsorption, and the nano-confined effect and low water saturation contribute to the enrichment of free gas in deep coalbed methane (CBM). These findings provide insights into the occurrence behavior of unconventional natural gas in reservoirs.
CHEMICAL ENGINEERING JOURNAL
(2023)
Article
Thermodynamics
Hu Wen, Li Yan, Yongfei Jin, Zhipeng Wang, Jun Guo, Jun Deng
Summary: Coalbed methane (CBM) disasters are a significant safety issue in coal mining, and CBM concentration prediction and early-warning technology are crucial in preventing and controlling these disasters. This study develops a fast and high-precision CBM concentration prediction method based on deep learning theory, using a large amount of coal mine CBM data. The proposed method combines multiple techniques, optimizes network parameters, and integrates with streaming technology to develop an efficient early-warning system that completes CBM concentration early-warning within 8 seconds. This method provides decision-making support for mine safety and CBM disaster prevention and control.
Article
Energy & Fuels
Dameng Liu, Yanbin Yao, Xuehao Yuan, Yanhui Yang
Summary: Field observations show that nearly 59% of CBM wells in the southern Qinshui Basin experience insufficient gas production due to the water-blocking effect caused by fracturing fluids or formation water. Experimental results indicate a stronger water-blocking effect in anthracite coal compared to bituminous coal, with water-blocking threshold pressure gradients following a positive power-law relationship with normalized water saturation. Study findings emphasize the importance of understanding gas and water interaction in coal reservoirs during gas production.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Engineering, Chemical
Jizhao Xu, Hexiang Xu, Cheng Zhai, Yuzhou Cong, Shuxun Sang, P. G. Ranjith, Quangui Li, Xiong Ding, Yong Sun, Yongshuai Lai
Summary: Low-field nuclear magnetic resonance (NMR) is used to analyze pore-size distributions in reservoirs, but the calculation of surface relaxivity (p) using conventional NMR methods is time-consuming and expensive. This study proposes an NMR grain packing method for rapid measurement of p values of unconsolidated coal cuttings. Different coal grain sizes exhibit different relaxation responses, influenced by grain packing patterns, grain morphology, water film distribution, and soluble minerals. The results show that macropores contribute significantly to porosity, and different models yield different p values for coal samples.
Article
Thermodynamics
Zixi Guo, Jinzhou Zhao, Zhenjiang You, Yongming Li, Shu Zhang, Yiyu Chen
Summary: Prediction of CBM production is crucial for its exploitation and utilization, with traditional methods limited. A new method based on deep learning shows higher accuracy by combining feature extraction for multiscale data sources and utilizing affinity propagation and LSTM network for prediction model.
Article
Thermodynamics
Shuyi Du, Jiulong Wang, Meizhu Wang, Jiaosheng Yang, Cong Zhang, Yang Zhao, Hongqing Song
Summary: This study develops an autonomous data governance system that combines supervised learning and unsupervised learning for anomaly detection and missing value supplement in coalbed methane (CBM) production data. A data-driven production forecasting model is also designed to tackle different production curve patterns of CBM wells.
Article
Energy & Fuels
Junqiang Kang, Derek Elsworth, Xuehai Fu, Shun Liang, Hao Chen
Summary: This study analyzed the impact of water saturation on the permeability of coal reservoirs through experiments and finite element methods, finding that changes in E and v can lead to a faster or slower decrease in permeability, with E having a more significant effect on permeability.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Geosciences, Multidisciplinary
Shasha Zhang, Caifang Wu, Xiaojie Fang, Ningning Liu, Xiuming Jiang, Jiang Han
Summary: The stress sensitivity and compressibility of the pore system in coal greatly affect coal permeability and porosity, which are crucial for coalbed methane development. Understanding water transformation under stress and during drying is important for analyzing water's impact on pore structure.
NATURAL RESOURCES RESEARCH
(2022)
Article
Computer Science, Interdisciplinary Applications
Zhibo Zhao, Yuhua Chen, Yi Zhang, Jinhui Luo, Guinan Mei, Heping Yan, Oluwasegun O. Onibudo
Summary: This study proposes a deep learning model, T-DGCN, for predicting complex gas production sequences by integrating time, space, and geological features. The model incorporates Dynamic Time Warping (DTW) to measure the similarity of geological features between wells and dynamically corrects the weight matrix in a multilayer neural network. Experimental results show that T-DGCN achieves an accuracy of 0.9298 in short-term production prediction, outperforming baseline models. Moreover, the use of DTW to calculate geological similarity significantly improves the model's performance.
COMPUTERS & GEOSCIENCES
(2023)
Article
Engineering, Geological
Shuai Chen, Linming Dou, Wu Cai, Lei Zhang, Miaomiao Tian, Zepeng Han
Summary: Coalbed methane (CBM) is an important unconventional fuel source, and its efficient extraction is crucial for reducing greenhouse gas emissions and ensuring coal mine safety. This study utilized micro-computed tomography and deep learning to analyze the behavior of coal fractures under liquid nitrogen cyclic fracturing. The results showed that LN2 treatment can effectively damage the coal sample and promote fracture expansion. The study provides insights into the application of LN2 cyclic fracturing in CBM recovery.
ROCK MECHANICS AND ROCK ENGINEERING
(2023)
Article
Automation & Control Systems
Christoeffer Lofflier, Christopher Mutschler
Summary: In this study, we propose IALE1, an imitation learning scheme for active learning, which can optimize the sample selection performance on different datasets and deep classifier architectures.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Automation & Control Systems
Christoffer Loeffler, Christopher Mutschler
Summary: This study proposes an imitation learning scheme for active learning, which optimizes the learning process by imitating the selection of the best expert heuristic. Experimental results show that the proposed scheme outperforms existing imitation learners and heuristics on image datasets.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Geosciences, Multidisciplinary
Taiyuan Zhang, Shuling Tang, Dazhen Tang, Shu Tao, Aobo Zhang, Yifan Pu
Summary: The anisotropy of coal pores and fractures (APF) has a significant impact on coal reservoirs, especially in high-dip coal seams. A low-field nuclear magnetic resonance (NMR) test can be used to detect the pore size distribution under different confining pressures and express the evolution of APF with stress. Pressurized NMR tests revealed that the majority of the micropores had geometric spindles parallel to the bedding plane, and the seepage fractures were mostly occupied by the cleat system perpendicular to the bedding plane. The compressibility of the fractures was related to the normal stress received.
NATURAL RESOURCES RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
F. Folino, G. Folino, M. Guarascio, F. S. Pisani, L. Pontieri
Summary: The study introduces a novel ensemble-based deep learning framework to address the non-stationary nature of IDS log data as well as the scarcity of labeled data instances, demonstrating effective classification accuracy and robustness to data scarcity in real-world scenarios.
INFORMATION FUSION
(2021)
Article
Energy & Fuels
Jianmeng Sun, Ping Feng, Peng Chi, Weichao Yan
Summary: Through experiments and numerical simulations, it is found that pore structure parameters of tight sandstone have a significant influence on the electrical properties. Constructing a saturation model can improve the calculation accuracy.
Article
Energy & Fuels
Naser Golsanami, Bin Gong, Sajjad Negahban
Summary: Gas-lift dual gradient drilling with oil-based drilling fluid is investigated in this study. Comparisons are made between different models to evaluate the PVT behavior and optimize the gas flow rate. The existing Standing model has errors in assessing the PVT behavior and is not recommended, while the new models provide more accurate results for mixture evaluation and optimization.
Article
Thermodynamics
Naser Golsanami, Madusanka N. Jayasuriya, Weichao Yan, Shanilka G. Fernando, Xuefeng Liu, Likai Cui, Xuepeng Zhang, Qamar Yasin, Huaimin Dong, Xu Dong
Summary: This study provides a detailed quantitative characterization of clays in hydrocarbon reservoirs using deep learning and SEM images, and investigates the impact of clays on reservoir fluid flow. The results show that clays can significantly reduce reservoir porosity and permeability, shedding new light on the detailed impacts of clay minerals on reservoir quality.
Article
Energy & Fuels
Bin Gong, Ruiqi Zhang, Tianwei Sun, Yujing Jiang, Naser Golsanami, Yanlong Li, Shanilka G. Fernando, Madusanka N. Jayasuriya
Summary: This study analyzes the deformation, slope stability, and sand production during methane hydrate exploitation through numerical simulations. The results show that increasing loading and decompression amplitudes lead to increased deformation degree and distribution, elevated pore water pressure, larger methane hydrate regeneration and dissociation areas. Additionally, sand production always starts from the upper and bottom sections of the methane hydrate-bearing sediments layer near the wellbore and increases with mining time.
Article
Energy & Fuels
Huaimin Dong, Jianmeng Sun, Muhammad Arif, Yihuai Zhang, Weichao Yan, Stefan Iglauer, Naser Golsanami
Summary: Gas hydrate reservoirs in the Muli area of China have complex lithology, high hardness, developed micro-fractures, and low porosity-permeability. This study combines digital rock technology with field-scale well-logging data to analyze the conductivity mechanism and identify the causes of the observed low resistivity. The findings provide fundamental information for accurate interpretation of key well-logging data and improving gas hydrate reservoir exploitation.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Green & Sustainable Science & Technology
Bin Gong, Ruijie Ye, Ruiqi Zhang, Naser Golsanami, Yujing Jiang, Dingrui Guo, Sajjad Negahban
Summary: Studying the failure mechanism of methane hydrate specimens is important for methane hydrate exploitation. Previous studies mainly focused on the macro or micromechanical response, while this study explored the mechanical response mechanism based on energy evolution. The numerical model of methane hydrate-bearing sediments was constructed and validated using laboratory tests. The simulation results qualitatively described the mechanical properties of the sediments and provided insights into deformation and failure mechanisms.
Article
Energy & Fuels
Huaimin Dong, Xin Zeng, Dalin Zhou, Jinjiang Zhu, Naser Golsanami, Jianmeng Sun, Yihuai Zhang
Summary: This study aims to characterize Langmuir's volume and Langmuir's pressure more effectively using well-logging data to evaluate the gas-bearing capacity of marine shales in the Wufeng-Longmaxi Formation in the southern Sichuan Basin, China. The influencing factors of shale conductivity were analyzed, and it was found that the factors affecting conductivity from strongest to weakest are conductive organic matter, thin low-resistivity layer, clay mineral, pore water, and pyrite.
JOURNAL OF ENERGY ENGINEERING
(2023)
Article
Engineering, Chemical
Suogui Shang, Qiangyong Gao, Yunjiang Cui, Peichun Wang, Zhang Zhang, Yadong Yuan, Weichao Yan, Peng Chi
Summary: We proposed a low-cost and high-efficiency workflow for simulating the electrical properties of rocks under high-pressure and high-temperature conditions. This method accurately predicts the electrical characteristics of rocks and plays a key role in water saturation prediction.
Article
Energy & Fuels
Qamar Yasin, Ali Gholami, Mariusz Majdanski, Bo Liu, Naser Golsanami
Summary: The geologic structure plays a decisive role in controlling fluid flow in geothermal systems. Seismic data can provide high-resolution images of complex structures, but it is challenging to detect the characteristics of fault edges in deep-buried and structurally complex areas. In this study, an intelligent workflow based on structure-oriented filtering and sensitive seismic attributes is proposed to characterize fault edges and fractures in geothermal reservoirs.
Article
Energy & Fuels
Behzad Saberali, Naser Golsanami, Kai Zhang, Bin Gong, Mehdi Ostadhassan
Summary: Intelligent surrogate models are important tools in reservoir simulation, and their efficiency depends on the quality of the data source. The current study uses a hybrid data source based on finite difference and streamline data, which speeds up database preparation and improves model training efficiency. The introduced surrogate model has been validated and shown satisfactory results, making it applicable in other reservoirs and gas exploitation from natural gas hydrate reservoirs in China.
GEOENERGY SCIENCE AND ENGINEERING
(2023)
Article
Mechanics
Behzad Saberali, Kai Zhang, Naser Golsanami
Summary: This study introduces a data-driven proxy modeling approach based on deep learning algorithms, which can determine the location of the injected water front in real-time. By minimizing the use of data extracted from numerical simulators and relying only on commonly available field data, the proposed proxy models successfully simulated breakthrough time and water arrival time in new blind scenarios.
INTERNATIONAL JOURNAL OF COMPUTATIONAL FLUID DYNAMICS
(2022)