Energy

Stone Ridge Technology Reduces Compositional Model Runtimes

Objective

Stone Ridge Technology and Eni released ECHELON 2.0, the first commercial reservoir simulation software designed from inception to run on NVIDIA GPUs with compositional simulation capabilities.

Customer

Stone Ridge Technology

Use Case

Simulation Software

Technology

NVIDIA A100 40GB Tensor Core GPUs, NVIDIA V100 Tensor Core GPUs

Since 2018, SRT and Eni have partnered from the initial deployment of ECHELON through ECHELON 2.0 to enhance software capabilities and deploy in industrial environments. Eni accelerated its digital transformation efforts with the HPC5 supercomputer consisting of 1,820 Dell EMC servers, each with four NVIDIA V100 Tensor Core GPUs.

Complex Reservoir Models Outpace Legacy Simulator Capabilities

Multi-core CPU simulators struggle to retain the reservoir detail and complexity needed for the accurate risk analysis of increasingly large geological models. Reservoir engineers spend days or weeks on large clusters simulating upscaled reservoir models.

“The speed and flexibility of ECHELON will be the key elements of the subsurface modelling in the energy transition pathway.”

Alessandro Caschili
Head of Reservoir Innovative Technologies, Eni

World’s Fastest Reservoir Simulation Software for Complex, Large-Scale Models

To address these issues, Stone Ridge Technology (SRT) and Eni developed ECHELON, the world’s fastest reservoir simulation software, to run multi-million cell models using the massive parallelism of NVIDIA GPUs. ECHELON gives reservoir engineers the opportunity to simulate fields using the scale of the geological grid, with the possibility to capture and better estimate uncertainty by running an ensemble of reservoir models.

ECHELON 2.0 is the most significant release of SRT’s reservoir simulation software since its introduction to the market in 2016. One of the key features is compositional formulation, which was planned, designed, implemented, and rigorously tested by SRT and Eni. ECHELON 2.0 was created to make full use of GPU capabilities and resources, optimizing for memory bandwidth, storage, and floating point operations per second (FLOPS). This release is the industry’s first compositional simulator exclusively formulated from inception to run on NVIDIA GPUs.

20-Year Simulation in 30 Minutes

To test ECHELON, Stone Ridge Technology made a goal to define 100 new well locations in a high-resolution geologic model using a fully implicit (FIM) simulation. ECHELON 2.0’s compositional formulation ran the 11.2-million cell model with nine components for 20 years of forecast. The entire simulation took 30 minutes with eight NVIDIA A100 40GB Tensor Core GPUs and 45 minutes with four NVIDIA A100 40GB Tensor Core GPUs. ECHELON’s FIM compositional formulation scales both runtime and memory requirements linearly with respect to the number of components, a unique feature in the energy industry.

Simulator Runtime

“ECHELON helps our team in better understanding the geological settings and fluid dynamics of one of the main strategic assets in Eni, a giant gas field located in North Africa. Its implementation significantly changed our activities because, thanks to the speed of ECHELON combined with the computational power of the HPC5 GPU cluster installed in the Eni Green Data Center, we were able to better calibrate the reservoir model introducing all the needed geological details and span all the relevant uncertainties by creating hundreds of simulations in a short time. This allowed us to not only fast track the project but also tailor the proper reservoir management strategy while weighting the associated risks to reach and maintain the expected gas plateau in due time.”

Marica Calabrese
Reservoir Studies North Africa and Middle East, Eni

Eni Accelerates Simulation Use Cases

In total, the system comprises 7,280 NVIDIA V100 GPUs. In a webinar presented by SRT and Eni, Eni presented results from three different types of use cases for ECHELON 2.0. The first was a deep-water reservoir using compositional formulation on four V100 GPUs for 150 runs on Eni’s HPC5 cluster.

In this trial, most realizations ran in less than four hours. The second was an offshore oil field that took 10 hours to run five reservoirs, each ranging from 300,000 to 2,100,000 grid blocks with multi-reservoir coupling. The third was a deep-water gas field with three reservoirs, each approximately 1,000,000 grid blocks, with reservoir coupling. In this third use case, SRT’s proprietary integrated-facility-network solver was linked to ECHELON and coupled at each time step. The full field development ran in 15 minutes.

Results

Launched first commercial compositional simulation software to run exclusively on NVIDIA GPUs

Ran an 11.2-million-cell model with nine components and forecast out 20 years in 30 minutes

Accelerated deep-water reservoir, offshore oil field, and deep-water gas field simulations with ECHELON 2.0 running on Eni’s HPC5 cluster

Setting a New Performance Bar for Reservoir Simulation Software

ECHELON’s compositional formulation sets a new bar for industry performance. Powered by NVIDIA GPUs, ECHELON opens up the possibility of simulating large systems with significantly more detailed fluid models. To accelerate future workloads, SRT is launching an ECHELON and NVIDIA DGX™ bundle, with ECHELON coming preinstalled in DGX for immediate deployment. Users can reduce their hardware footprint for large-scale simulations by replacing thousands of processors in a data center that covers half a football field with a compact bundle that fits in the space of half a table tennis table.