Pros and Cons of Applying a Proxy Model as a Substitute for Full Reservoir Simulations
- Dennis Denney (JPT Senior Technology Editor)
- Document ID
- Society of Petroleum Engineers
- Journal of Petroleum Technology
- Publication Date
- July 2010
- Document Type
- Journal Paper
- 41 - 42
- 2010. Society of Petroleum Engineers
- 10 in the last 30 days
- 257 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||Free|
|SPE Non-Member Price:||USD 15.00|
This article, written by Senior Technology Editor Dennis Denney, contains highlights of paper SPE 124815, "Pros and Cons of Applying Proxy Models as a Substitute for Full Reservoir Simulations," by D.I. Zubarev, SPE, SPT Group, prepared for the 2009 SPE Annual Technical Conference and Exhibition, New Orleans, 4-7 October. The paper has not been peer reviewed.
A comparative study was made of proxy-modeling methods (also known as surrogate modeling or metamodeling) as a computationally inexpensive alternative to full numerical simulation in assisted history matching, production optimization, and forecasting. The study demonstrated the solution-space complexity for different simulation models and the applicability of the proxy models to mimic it. Focus was given to practical aspects of model construction and to limitations of which engineers should be aware.
Recent improvements in computational hardware and software have advanced reservoir modeling. However, for many workflows in uncertainty quantification and optimization with application to reservoir simulation, the availability of computing resources is still a limiting factor.
Here, a “proxy model” is a mathematically or statistically defined function that replicates the simulation-model output for selected input parameters. Typical proxy models in reservoir simulation include the following.
- Sensitivity analysis of uncertainty variables
- Probabilistic forecasting and risk analysis
- Conditioning a simulation model to observed data (history matching)
- Field-development planning and production optimization
Proxy models, combined with design-of-experiment techniques, are used widely for sensitivity analysis. Application scenarios include the traditional one-parameter-at-a-time approach for linear-sensitivity analyses and advanced experimental designs that are capable of resolving correlation and higher-order effects. For probabilistic forecasting, proxy models are used routinely as input to a Monte Carlo sampling process. The high computational efficiency of proxy models enables exhaustive sampling rates.
This study investigated the predictive quality and computational effort required for different proxy-modeling algorithms to provide acceptable results in solving reservoir-simulation problems. Three simulation models of different complexities were selected to understand the effect of the simulation-model structure on the efficiency of the proxy-model performance. Different application workflows [i.e., history matching, production optimization, hydrocarbons initially in place (HCIIP), and oil-recovery forecasting] were used to investigate the effect of the simulation-model output-data type on the proxy-model performance.
|File Size||75 KB||Number of Pages||2|