A5101899852- Enhanced Oil Recovery Techniques
- Hydraulic Fracturing and Reservoir Analysis
- Hydrocarbon exploration and reservoir analysis
- CO2 Sequestration and Geologic Interactions
- Domain Adaptation and Few-Shot Learning
- Color Science and Applications
- Reservoir Engineering and Simulation Methods
- Color perception and design
- Petroleum Processing and Analysis
- Machine Learning and ELM
- Multimodal Machine Learning Applications
- Atmospheric and Environmental Gas Dynamics
- Visual perception and processing mechanisms
- Advanced Chemical Sensor Technologies
- Photonic and Optical Devices
- Coal Properties and Utilization
- Image Enhancement Techniques
- Sparse and Compressive Sensing Techniques
- Geological Modeling and Analysis
- Advanced Technologies in Various Fields
- Photonic Crystals and Applications
- Advanced Data Compression Techniques
- Semiconductor Lasers and Optical Devices
- Carbon Dioxide Capture Technologies
- Data-Driven Disease Surveillance
China University of Petroleum, Beijing
2021-2024
Beijing Institute of Graphic Communication
2022-2024
Xi'an Jiaotong University
2023-2024
Sun Yat-sen University
2024
Beijing University of Technology
2023-2024
Nantong University
2023
Oil and Gas Center
2022
Zhejiang International Studies University
2022
Driven by practical needs, research on Class-Incremental Learning (CIL) has received more and attentions in recent years. A technical challenge to be conquered CIL methods is the catastrophic forgetting problem, where model's performance improves rapidly new classes while deteriorates drastically old ones. The main causes behind include network drifts, inter-class confusions, etc. In this paper, we propose a novel method that solves problem from two aspects. First, solve confusion Semantic...
Carbon capture, utilization, and storage (CCUS) is a green engineering technology to reduce CO2 emissions mitigate climate warming. It crucial accurately predict the CO2–brine interfacial tension (IFT) in order evaluate carbon capacity of saline aquifers. Traditional experimental methods are time-consuming costly. The existing empirical correlation IFT have been found be inaccurate. Instead, machine learning (ML) superior ability IFT. However, lack an in-depth examination main factors...
Driven by concerns over carbon neutrality, enhancing gas recovery coupled with CO2 sequestration (CO2-EGR) has emerged as a prominent research topic worldwide. Tight sandstone reservoirs are characterized low primary rate, and the adsorbed CH4 within reservoir is challenging to recover via conventional pressure decay. In this study, we first classified tight based on their pore-throat structural characteristics. Subsequently, conducted isothermal adsorption experiments, relative permeability...
As a means of delaying climate change, injecting carbon dioxide (CO2) into geological structures can be an effective capture, utilization, and storage (CCUS) strategy. All are naturally spatial heterogeneity, which significantly affect fluid flow heat transfer, thereby affecting CO2 storage. We modeled transfer in 12 scenarios with spatially heterogeneous reservoirs using thermal-hydraulic-mechanical (THM) coupled model 3D wellbore-reservoir system. injectability distribution uniformity were...
In recent years, use of photonic crystals has been recognized as a viable approach to resolve the contradiction between operating speed and light absorption in vertical-structure photodiodes. this paper, we present crystal that enhances responsivity GeSn pin photodiode near-infrared region spectrum. Additionally, effects on most important properties photodiodes are studied. Based measurement analysis device's dark current, responsivity, characteristics, structure reduces current by...
In the context of incremental class learning, deep neural networks are prone to catastrophic forgetting, where accuracy old classes declines substantially as new knowledge is learned. While recent studies have sought address this issue, most approaches suffer from either stability-plasticity dilemma or excessive computational and parameter requirements. To tackle these challenges, we propose a novel framework, Diverse Knowledge Transfer Transformer (DKT), which incorporates two transfer...
The deterioration in lake water environments, especially increasing eutrophication, is prevalent all over the world, which has seriously affected balance and stability of internal ecosystem lakes. In this study, modern sediment samples were collected from three subtropical freshwater lakes with significant differences nutrient levels to analyze concentration zooplankton Cladocera Bosminidae its relationship lakes’ ecological changes. results show that environments caused by eutrophication...
Carbon dioxide-enhanced oil recovery (CO2-EOR) is the main application of carbon capture, utilization, and storage (CCUS) in gas field development, modeling evaluation process minimum miscible pressure (MMP) CO2–crude system very important for CO2-EOR projects. Unlike previous studies that often used low-dimensional incomplete data, this research utilizes high-dimensional nonlinear full-component tabular data injected crude oil. The mixed screening method employed to identify significant...
Both observer age and size of stimulus as characterized using the field view (FOV) are two important parameters to affect color matching functions (CMFs) human observers. They also included in cone fundamental CMFs models that were recently proposed by International Commission on Illumination (CIE) 2006. In contrast great number studies investigating performance characterizing matches mismatches different primary sets, few study investigated effect these factors. this study, we carefully...
The primary application of carbon capture, utilization and storage (CCUS) in oil gas field development engineering is CO2 enhanced recovery (CO2-EOR). However, how to systematically flexibly evaluate the potential CO2-EOR technology still a major challenge. Traditional evaluation methods are mostly user-defined weights, which have great subjective limitations. In addition , missing unbalanced oilfield data further limits key technologies development. this study, new method based on hybrid...
Existing prompt learning methods in Vision-Language Models (VLM) have effectively enhanced the transfer capability of VLM to downstream tasks, but they suffer from a significant decline generalization due severe overfitting. To address this issue, we propose framework named LOBG for vision-language models. Specifically, use CLIP filter out fine-grained foreground information that might cause overfitting, thereby guiding prompts with basic visual concepts. further mitigate devel oped...