Kabosu: ECHELON Software Tackles Complex Field with Flying Colors

ECHELON software tackles an extremely difficult compositional model. A demonstration of the robustness of ECHELON software.

Bg case study kabosu

Mar 08 2022

A demonstration of the robustness of ECHELON software.

By the numbers
0
Wells
0
Years of Forecast
0
Components
Stone Ridge Technology – ECHELON Advantages

Challenge

The Kabosu field is as complex as it gets to simulate. It combines a challenging dual permeability grid with 1.4M active cells and local grid refinement around the injectors, gas injection at miscible conditions, and complex field management strategies. The challenge was to achieve fast enough runtimes to perform sensitivity studies on an ensemble of models within the same working day, without sacrificing accuracy.

Solution

One of the key features of ECHELON is a compositional formulation created from inception to make full use of the GPUs’ capabilities and resources, optimizing for memory bandwidth, storage and FLOPS; it has been planned, designed, implemented and rigorously tested by the teams at SRT and Eni.

ECHELON is commercialized with fully implicit (FIM) and adaptive implicit (AIM) formulations. It supports most reservoir and field management options available in other reservoir simulators, with the exceptional performance you expect from ECHELON. The software uses industry-standard input and output formats, and is therefore compatible with well-known pre and post-processing tools; adoption is a seamless task.

Results

The Kabosu model was run on compute nodes with 4x V100 GPUs and 2x Intel 4114 silver CPUs (10 cores each). Results using 1, 2, 3 and 4 GPUs are presented below. Model turn-around in approximately 2 hours allows multiple engineering iterations within a workday, enhancing productivity.

Results – Stone Ridge Technology

Figure 1: Elapsed time vs. simulated time for the Kabosu model ran with ECHELON on 1, 2, 3, and 4 NVIDIA V100 GPUs.

The cumulative number of Newton and linear iterations as well as the number of wells producing at a given point in time is essentially unaffected by the GPU count. The tail period is the most challenging part of the forecast as most wells are on THP control and producing below the bubble-point. This further confirms the robustness of ECHELON software.

Case study kabosu results

Figure 2: Cumulative number of Newton and linear iterations, and number of flowing producer wells vs. simulated time, for the Kabosu model ran with ECHELON on 1, 2, 3, and 4 NVIDIA V100 GPUs.

Panfili, P., Cominelli, A., Calabrese, M., Albertini, C., Savitskiy, A., and Leoni, G. (2012). Advanced upscaling for Kabosu reservoir modeling. SPE Reservoir Evaluation and Engineering, 15(02):150–164. SPE-146508-PA.

We thank Eni S.p.A for the permission to publish the data contained in this case study.

Additional case studies

Deep dives into our software product and its efficacy.