EI、Scopus 收录
中文核心期刊
宋洪庆, 都书一, 王九龙, 劳浚铭, 谢驰宇. 数智流体力学的发展及油气渗流领域应用. 力学学报, 2023, 55(3): 765-791. DOI: 10.6052/0459-1879-22-484
引用本文: 宋洪庆, 都书一, 王九龙, 劳浚铭, 谢驰宇. 数智流体力学的发展及油气渗流领域应用. 力学学报, 2023, 55(3): 765-791. DOI: 10.6052/0459-1879-22-484
Song Hongqing, Du Shuyi, Wang Jiulong, Lao Junming, Xie Chiyu. Development of digital intelligence fluid dynamics and applications in the oil & gas seepage fields. Chinese Journal of Theoretical and Applied Mechanics, 2023, 55(3): 765-791. DOI: 10.6052/0459-1879-22-484
Citation: Song Hongqing, Du Shuyi, Wang Jiulong, Lao Junming, Xie Chiyu. Development of digital intelligence fluid dynamics and applications in the oil & gas seepage fields. Chinese Journal of Theoretical and Applied Mechanics, 2023, 55(3): 765-791. DOI: 10.6052/0459-1879-22-484

数智流体力学的发展及油气渗流领域应用

DEVELOPMENT OF DIGITAL INTELLIGENCE FLUID DYNAMICS AND APPLICATIONS IN THE OIL & GAS SEEPAGE FIELDS

  • 摘要: 大数据及人工智能技术的崛起推动了数智流体力学的快速发展. 数智流体力学是将流体力学、大数据和人工智能相结合, 以流体力学场景需求为导向, 形成以“数”为基础, 以“智”为核心, 以算力为支撑的新研究范式. 核心内涵是要以数据驱动为主, 融合物理信息、专家经验等先验知识, 利用智能化手段构建“数据 + 物理”双驱动的数智模型, 解决场景需求问题. 数智流体力学在建模灵活性、运算效率、计算精度方面具有十分明显的优势, 其应用潜力已经在多尺度流动、多场耦合以及流场建模等方面得到验证. 数智流体力学研究范式包括数据治理和智能算法构建, 其中数据治理工作尤为重要, 治理后的数据质量是智能算法能否发挥其价值的关键. 智能算法中“数据 + 物理”协同驱动主要存在四种引入机制, 分别是基于输入数据的嵌入机制、基于模型架构的嵌入机制、基于损失函数的嵌入机制和基于模型优化的嵌入机制. 以油气领域应用为例, 介绍了数智流体力学在储层物性参数预测、压裂效果评价以及注采参数优化等方面的一系列研究进展. 数智流体力学是流体力学未来的重要发展方向之一, 以场景需求为导向、深度融合物理信息等先验知识的新一代智能理论与方法是数智流体力学发展的必然趋势, 能够从崭新的角度攻克诸多复杂多变的流体力学关键问题.

     

    Abstract: The growth of big data and artificial intelligence technologies has driven the rapid development of digital intelligence fluid mechanics. Digital intelligence fluid mechanics combines fluid mechanics, big data and artificial intelligence, to establish a new research paradigm oriented to specific scenarios of fluid mechanics, with "data" as the basis, "intelligence" as the core, and arithmetic power as the support. Its connotation is to establish a "data + physics" co-driven digital intelligence model, which is mainly data-driven and incorporates prior knowledge such as physical information and expert experience, to solve practical problems in different scenarios. Digital intelligence fluid dynamics has very obvious advantages in modeling flexibility, computing efficiency, and computational accuracy, whose application potential has been proven in multi-scale flow, multi-field coupling, and flow field modeling. In terms of the construction of digital intelligence models, data governance is indispensable since the data quality improved by governance enables intelligent algorithms to perform preferably. There are four main mechanisms for introducing "data + physics" co-driving in intelligent algorithms, which are input data-based embedding mechanism, model architecture-based embedding mechanism, loss function-based embedding mechanism and model optimization-based embedding mechanism. Taking oil & gas field applications as an example, a series of research advances in the prediction of physical parameters, evaluation of fracturing effects and optimization of injection parameters by digital intelligence fluid dynamics are introduced. Future diversified research models can take advantage of the efficient and rapid modeling of digital intelligence fluid dynamics, but also ensure physical interpretability and extrapolation in both classical and computational fluid dynamics. Therefore, digital intelligence fluid mechanics is an inevitable trend in the future development of fluid mechanics, and it is necessary to take the scenario demand as the guide, deeply integrate physical information and prior knowledge, actively explore new intelligent theories and methods, and attack the complex and changing scientific problems in fluid mechanics.

     

/

返回文章
返回