Advertisement

Figure 1: Oil deposited inside the rock environment or the reservoir.3 Contributed by Dr. Zhi Shang, Post-Doctoral Researcher, Intel PCC at LSUThis article looks at the impact of high performance computing on the reservoir modeling portion of the oil and gas exploration workflow. It examines how the Intel Parallel Computing Center (Intel PCC) at the LSU Center for Computation & Technology uses supercomputers and HPC tools for micropore analysis of oil and gas reservoirs. In addition, the center studies sand accumulations in porous media to determine how to define an oil and gas extraction path to avoid sand clogs in well-boring equipment and other facility equipment.

Introduction to oil and gas reservoirs

In the oil and gas industry, gas, oil, water and sand particles all normally exist in a reservoir. Oil and gas companies need accurate information on the location of oil within porous media (such as rock and sand) to help them determine where to drill and to evaluate existing oil reservoirs. It is also important to know where extensive sand deposits exist within porous media.

Research at the Intel PCC at LSU is helping to determine how to organize and control the flow fields inside the porous media so that sand particles do not block fluid transport. This type of research requires the use of supercomputers and HPC tools due to the large number of flow regimes and large size domains that have to be studied. In addition, the flow regimes have to be wide enough to cover both laminar flow and turbulent flow.

In the process of oil extraction, multiphase flows (gas, oil, water and sand particles) will be extracted into the oil and gas extraction equipment.1 This introduces problems for oil and gas facilities — a major issue is the blockage of the oil transportation. This block frequently happens between the oil and gas well casing and the reservoir. For example, the dispersion of solid particles inside the porous media will affect the rock environment around a well casing and the reservoir. Therefore, an understanding of fluid mechanics in general multiphase flow distribution through porous media should have a significant impact on the well productivity, well storage capacity, production costs and equipment design. Many engineers and researchers employ computational fluid dynamics (CFD), to study the details of these issues to design their facilities.2

Micropore research

The oil and gas industry understands the importance of computer simulation and HPC technology to provide information on the distribution of various media in reservoirs. The Intel Parallel Computing Center (Intel PCC) at the LSU Center for Computation & Technology (CCT) is using supercomputers and HPC software tools to study the flow through porous media, which is fundamental to many engineering and environmental processes related to oil and gas, chemical processing, hydrology and geophysics. A special area of exploration at LSU is studying the simulation of flows through micropores, such as those found in rocks involved in oil and gas extraction. Simulation challenges are magnified by the complicated geometry of void spaces in rocks or packed bed reactors. Recent advances in HPC capabilities, such as high-fidelity mesh generation and flow solvers for these computational domains, have made image-based pore-scale modeling a viable technique for addressing such problems.

The Intel PCC at LSU uses the Eulerian-Lagrangian approach based on DPM (discrete particle modeling) in OpenFOAM to study the distribution and dispersion of solid particles inside the porous media of a natural dual-permeability rock.

Figure 1 shows the oil deposited inside the rock environment of the reservoir. Dr. Zhi Shang, Post-Doctoral Researcher, Intel PCC at LSU explains, “The pore geometries of the porous media were generated by sphere packing. The porosity is about 37 percent, which is approaching the natural dual-permeability of rock. For this study, a series of simulation cases were established from small scale of 1 million mesh cells with 1.2 million sand particles to large scale 50 million mesh cells with 20 million sand particles. The sand particles are distributed by Gaussian distribution from 20 μm to 180 μm with a mean value of 100 μm.”

Figure 2: Distribution of sand particles inside the porous media4 Contributed by Dr. Zhi Shang, Post-Doctoral Researcher, Intel PCC at LSU The team also studies the flow behaviors of large-scale solid suspension flows in porous media. Through the simulations, it can be seen that, inside the porous media, the sand particles accumulate in some areas, shown in Figure 2. The sand accumulation can increase the risk of blocking the sub-channels for fluid transport.

SuperMIC supercomputer and HPC software

For the large-scale simulations in this research, SuperMIC (LSU’s newest 1-PF class Intel Xeon Phi coprocessor equipped cluster), serves as the development platform. SuperMIC is capable of a peak theoretical performance of over 925 TF. It achieved a performance of 557 TF during testing, which placed it as number 65 in the June 2014 Top500 List.

Figure 3: Intel Xeon processor and Intel Xeon Phi coprocessor MIC architecture.6 Contributed by Dr. Zhi Shang, Post-Doctoral Researcher, Intel PCC at LSUProcessing is done using Intel Xeon processors and Intel Xeon Phi co-processors; the architecture is shown in Figure 3, where it can be seen that Intel Xeon Phi coprocessors can be the auxiliary expansion device for large-scale computing.

For performing the CFD simulations using OpenFOAM, the Intel Parallel Studio XE Cluster Edition 2015 Version 5.0.1.035 MPI library and compiler were used for the parallel processing on the SuperMIC cluster. The open source Lustre file system version 2.6 was on the Intel Xeon Phi coprocessor system that was used for data management. The work disk was mounted by NFS/Lustre on Intel Xeon Phi coprocessors. Therefore, the file system can be shared between the Intel Xeon processor host and the Intel Xeon Phi coprocessors directly through the hardware.

According to Dr. Shang, “Using the hardware of the Intel Xeon processor and Intel Xeon Phi coprocessors extends the resource of the supercomputer. With the help of Intel Xeon Phi coprocessors, parallel computing was upgraded. For example, with one Intel Xeon Phi coprocessor 7120P on the SuperMIC cluster, the MPI can be enlarged by an extra 61 tasks. This is equivalent to upgrading a PC to a workstation and a workstation to a cluster. This is really helpful for the research we are doing.”

Figure 4: Speed up of symmetric MPI mode on Intel Xeon and Intel Xeon Phi coprocessors. Contributed by Dr. Zhi Shang, Post-Doctoral Researcher, Intel PCC at LSUThe LSU SuperMIC cluster provides a powerful platform for large-scale scientific and engineering computing for the team’s research. “Due to the expansion of Intel Xeon Phi coprocessors, the porous media simulation is able to run on hybrid Intel Xeon processors and Intel Xeon Phi coprocessors based on MPI in the SuperMIC cluster. Figure 4 shows the speed up curves of the pure MPI running on Intel Xeon processors and hybrid Intel Xeon processors and Intel Xeon Phi coprocessors. The symmetric running mode was chosen for the hybrid Intel Xeon processors and Intel Xeon Phi coprocessors. During parallel running, the MPI tasks on Intel Xeon Phi coprocessors occupied approximately 30 percent average of the entire MPI tasks,” states Dr. Shang.

Recommendations for future work

“From our oil and gas industry research, our team found that using HPC software and Intel Xeon Phi coprocessors can help organizations apply new technologies and processes that will capture and transform raw data into actionable insights. This helps the oil and gas industry improve asset value and yield while enhancing safety and protecting the environment. In our research, well and field operations are instrumented to capture a holistic view of equipment performance and well productivity data including reservoir, well, facilities and export data.7 I believe that, in the future, the oil and gas industry will increasingly need to use large-scale simulations and big data processing for their designs and operations,” states Dr. Shang.

The final stage of oil and gas exploration workflow is Financial Analysis. With the data collected, and the imaging, interpretation and modeling provided in the previous stages, industry executives are in the position to make smart decisions about their deployment of resources.

Acknowledgments

Dr. Shang would like to thank the support by Intel PCC at LSU - Intel Parallel Computing Center at Louisiana State University, LA 70808, USA (LSU# Y1SY1-1) and portions of this research were conducted with high performance computing resources at SuperMIC provided by Louisiana State University.

References

  1. Zhi Shang, Jing Lou, Hongying Li. CFD of transition from bubbly flow to slug flow in vertical pipe. International Journal of Chemical Engineering and Processing, 1(1), February 2015, 14~20.
  2. Zhi Shang. Performance analysis of large scale parallel CFD computing based on Code_Saturne. Computer Physics Communications, 184 (2), February 2013, 381~386.
  3. http://resources.schoolscience.co.uk/ExxonMobil/infobank/4/2index.htm?origin.html.
  4. Zhi Shang, Honggao Liu, Krishnaswamy Nandakumar, Mayank Tyagi, James A. Lupo, Karten Thompson. High Performance Computing at Intel Xeon Phi Coprocessor Using Native and Symmetric Modes for Discrete Particle Model of OpenFOAM. 3rd Annual EPIC Workshop on Enabling Process Innovation through Computation, Louisiana State University, Baton Rouge, Louisiana, May 1, 2015, USA.
  5. http://www.hpc.lsu.edu/resources/hpc/system.php?system=SuperMIC.
  6. http://colfaxresearch.com/configuration-and-benchmarks-of-peer-to-peer-communication-over-gigabit-ethernet-and-infiniband-in-a-cluster-with-intel-xeon-phi-coprocessors.
  7. http://www-01.ibm.com/software/data/bigdata/industry-oil.html.

Linda Barney is the founder and owner of Barney and Associates, a technical/marketing writing, training and web design firm in Beaverton, OR.

Advertisement
Advertisement