VTechWorks staff will be away for the Thanksgiving holiday beginning at noon on Wednesday, November 27, through Friday, November 29. We will resume normal operations on Monday, December 2. Thank you for your patience.

Browsing by Author "Zhang, Kaiyi"

Now showing 1 - 3 of 3
  • Thumbnail Image
    CO2 Minimum Miscibility Pressure and Recovery Mechanisms in Heterogeneous Low Permeability Reservoirs
    Zhang, Kaiyi (Virginia Tech, 2019-09-16)
    Benefited from the efficiency of hydraulic fracturing and horizon drilling, the production of unconventional oil and gas resources, such as shale gas and tight oil, has grown quickly in 21th century and contributed to the North America oil and gas production. Although the new enhancing oil recover (EOR) technologies and strong demand spike the production of unconventional resources, there are still unknowns in recovery mechanisms and phase behavior in tight rock reservoirs. In such environment, the phase behavior is altered by high capillary pressure owing to the nanoscale pore throats of shale rocks and it may also influence minimum miscibility pressure (MMP), which is an important parameter controlling gas floods for CO2 injection EOR. To investigate this influence, flash calculation is modified with considering capillary pressure and this work implements three different method to calculate MMP: method of characteristics (MOC); multiple mixing cell (MMC); and slim-tube simulation. The results show that CO2 minimum miscibility pressure in nanopore size reservoirs are affected by gas-oil capillary pressure owing to the alternation of key tie lines in displacement. The values of CO2-MMP from three different methods match well. Moreover, in tight rock reservoirs, the heterogeneous pore size distribution, such as the ones seen in fractured reservoirs, may affect the recovery mechanisms and MMP. This work also investigates the effect of pore size heterogeneity on multicomponent multiphase hydrocarbon fluid composition distribution and its subsequent influence on mass transfer through shale nanopores. According to the simulation results, compositional gradient forms in heterogeneous nanopores of tight reservoirs because oil and gas phase compositions depend on the pore size. Considering that permeability is small in tight rocks and shales, we expect that mass transfer within heterogeneous pore size porous media to be diffusion-dominated. Our results imply that there can be a selective matrix-fracture component mass transfer during both primary production and gas injection secondary recovery in fractured shale rocks. Therefore, molecular diffusion should not be neglected from mass transfer equations for simulations of gas injection EOR or primary recovery of heterogeneous shale reservoirs with pore size distribution.
  • Thumbnail Image
    Effect of Pore Size Heterogeneity on Hydrocarbon Fluid Distribution, Transport, and Primary and Secondary Recovery in Nano-Porous Media
    Zhang, Kaiyi; Du, Fengshuang; Nojabaei, Bahareh (MDPI, 2020-04-03)
    In this paper, we investigate the effect of pore size heterogeneity on fluid composition distribution of multicomponent-multiphase hydrocarbons and its subsequent influence on mass transfer in shale nanopores. The change of multi-contact minimum miscibility pressure (MMP) in heterogeneous nanopores was investigated. We used a compositional simulation model with a modified flash calculation, which considers the effect of large gas–oil capillary pressure on phase behavior. Different average pore sizes for different segments of the computational domain were considered and the effect of the resulting heterogeneity on phase change, composition distributions, and production was investigated. A two-dimensional formulation was considered here for the application of matrix–fracture cross-mass transfer and the rock matrix can also consist of different segments with different average pore sizes. Both convection and molecular diffusion terms were included in the mass balance equations, and different reservoir fluids such as ternary mixture syntactic oil, Bakken oil, and Marcellus shale condensate were considered. The simulation results indicate that oil and gas phase compositions vary in different pore sizes, resulting in a concentration gradient between the two adjacent pores of different sizes. Given that shale permeability is extremely small, we expect the mass transfer between the two sections of the reservoir/core with two distinct average pore sizes to be diffusion-dominated. This observation implies that there can be a selective matrix–fracture component mass transfer as a result of confinement-dependent phase behavior. Therefore, the molecular diffusion term should be always included in the mass transfer equations, for both primary and gas injection enhanced oil recovery (EOR) simulation of heterogeneous shale reservoirs.
  • Thumbnail Image
    Scalable Combinatorial Algorithms for Optimal Transport Based Similarity Metrics
    Zhang, Kaiyi (Virginia Tech, 2024-11-22)
    Optimal Transport (OT), also known as Wasserstein distance, is a valuable metric for comparing probability distributions. Owing to its appealing statistical properties, researchers in various fields, such as machine learning, use OT within applications. However, computing both exact and approximate OT is computationally expensive and impractical for large datasets. Furthermore, OT is sensitive to small noise in the input distributions. In this document, we propose to use combinatorial methods to design scalable and noise-resistant solutions for OT. We present four key contributions in this work. First, we introduce a novel combinatorial parallel algorithm for approximating OT, which achieves a parallel time complexity of $O(log n/varepsilon^2)$, where $n$ is the input size and $varepsilon$ is the addtitive error, Our algorithm outperforms the state-of-the-art in experiments. Second, we propose a new concept, OT-profile, representing the function of minimum partial optimal transport cost $sigma_alpha$ versus the transported mass $alpha$. This can be used to identify outliers in real-world data. The utility of OT-profile is demonstrated in outlier detection and PU-learning jobs and outperforms the state-of-the-art. Third, building upon the OT-profile, we propose a new OT-based metric for comparing distributions that is more robust to noise. This metric preserves desirable properties while reducing its sensitivity to noise for high $p$ values, providing a robust solution for real-world datasets. Lastly, we have developed a Python library that integrates our algorithms and methods into a user-friendly framework, making it easier for practitioners to adopt our methods. Our work enhances the computational efficiency and robustness of OT, making it practical for machine learning applicaitons.