期刊
FUEL
卷 333, 期 -, 页码 -出版社
ELSEVIER SCI LTD
DOI: 10.1016/j.fuel.2022.126296
关键词
Carbon capture and storage; Artificial intelligence; Machine learning; Forecasting technique; Geological sequestration
Greenhouse gas emissions cause global climate change and curbing CO2 emissions is urgent. Carbon capture and storage (CCS) is considered as an important choice to mitigate the greenhouse effect and machine learning (ML) has the potential to play a crucial role in CCS, such as predicting physical properties, evaluating stability, and monitoring CO2 migration and leakage. This review provides a comprehensive overview of ML applications in CCS, highlighting the challenges and future prospects.
Greenhouse gas emissions cause serious global climate change, and it is urgent to curb CO2 emissions. As the last -guaranteed technology to reduce carbon emissions, carbon capture and storage (CCS) is emerging and has the potential to become an important choice to mitigate the greenhouse effect in the future. Forward-looking project deployment and accurate forecasting techniques occupy an indispensable position in CCS. Machine learning (ML), one of the fastest developing intelligent technology fields at present era, is considered as a substantial means to realize forecast demand relying on computer science and data statistics. This work provided a comprehensive review of ML applications in CCS, based on classical ML methods and mainstream research di-rections in CCS. The study shown that ML algorithms such as artificial neural network (ANN) and convolutional neural network (CNN) were widely used, mainly for predicting physical properties, evaluating mechanical sta-bility, and monitoring CO2 plume migration and leakage during CO2 storage. Support vector machine (SVM) was generally combined with other ML methods for the prediction of petrophysical properties and sensitivity analysis of influencing factors. Deep learning (DL) algorithms, represented by generative adversarial network (GAN) and long short-term memory (LSTM), had shown good results in real-time monitoring of CO2 migration and leakage. Decision tree (DT) and random forest (RF) were mainly used to establish risk assessment and decision analysis framework, and estimate the success probability of CCS. This review summarized the applications of ML algo-rithms in CCS and presented the challenges and future prospects, from the geoscience perspective. The findings of this work can help for better understanding of the key role played by ML in CCS, as well as guiding the selection of ML algorithms and the development of new models in CCS research.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据