Reservoir Simulation Ensemble Analysis via CoViz 4D Visualization and Jupyter Notebook Calculations

Introduction to Scripting with CoViz 4D  

Dynamic Graphics provides powerful, interactive functionality to simplify software usage, but it has always recognized the need to prevent the limitations of a “black box” environment, allowing for the extension and customization of the tools by providing users with access to the underlying modular functionality. This enables clients to build their own solutions for precise problem solving, analysis, and 4D visualization without having to navigate the existing buttons and toolbars. CoViz 4D offers an extensive array of command line modules accessible via bash scripting as well as through the Python Developers’ Toolkit.

DGI’s efforts in this area are continuously advancing and we have developed further functionality in a Python package, ‘dgpy’ which reduces the barriers to entry in converting bash commands to Python. Additionally, we have expanded our current command modules, notably ‘cv_viewserver’, to include more of the options available in the CoViz 4D Viewer GUI. This progress has led to increasingly powerful, flexible, and efficient new tools that take advantage of programming with our Developers’ Toolkits, such as Cloud-computing and writing Python Jupyter notebooks that access DGI modules.

Use Case: Ensemble Analysis and Oil Thickness Calculations via a Jupyter Notebook

The use of ensembles of data, often involving tens, hundreds, or thousands of reservoir simulation models, is one approach toward better understanding and incorporation of reservoir uncertainty in model construction and reservoir performance predictions. An ensemble represents the range of uncertainty in a model – sources of uncertainty include lack of information in areas where there is no well data, bad data, imprecise information, and other factors. CoViz 4D contains tools for working with ensemble data to simplify the analysis of large data volumes and output the results as maps and images which are easy to visualize and aid understanding.

An offshore team needed to analyze an ensemble dataset, operating on, and comparing dozens of reservoir simulation models. These simulation models shared the same geospatial coordinates, wells, and properties, but the property data varied from model to model based on a Monte Carlo simulation. To conduct an in-depth analysis, individual model comparisons were necessary even though many of the same calculations were being run. Using a graphical user interface would have been slow and inefficient, and writing a generalized script would not have allowed for individual choices in comparisons. This challenge presented the perfect opportunity for the team to access CoViz 4D’s command line modules from within a Python Jupyter notebook.

The team’s goals were two-fold: 1) to compare a user-selected attribute at a user-selected timestep across every model in the ensemble and, 2) to calculate the difference in oil thickness between two user-selected models in the ensemble, and to compute and overlay the difference in Cumulative Oil Production for each well in the two models. In both cases the team needed a solution where the inputs/outputs could be easily changed, and new results generated rapidly.

The CoViz 4D ensemble summing module (cv_vdbensemblesum) outputs a single simulation model containing attributes representing the mean, minimum, maximum, and standard deviation across an ensemble of Nexus VDBs. In this particular use case, calculating and visualizing all four properties for every timestep in the ensemble was superfluous to the team’s requirements. Instead, the cv_vdbensemblesum module was accessed from within a custom Python Jupyter notebook where a single timestep and a specific attribute (e.g., SO_max) could easily be selected and calculated on the fly. The resulting model and attribute were loaded to a CoViz viewer within the notebook for easy QC.

Figure 1. .SO_max computed for an ensemble of 20 reservoir simulation grids.

For the second goal, determining the difference in oil thickness between two models in the ensemble, a custom Python Jupyter notebook was utilized to do the following steps:

  1. Read two user-selected simulation grids from the ensemble.
  2. Compute oil thickness (porosity*Oil Saturation*Net to Gross) at a user-given timestep for each model.
  3. For each model, sum the oil thickness over a user input k-layer range, then calculate the difference and output the result as a 2D map.
  4. Read the Cumulative Oil Production (COP) data for all wells in the two simulation models.
  5. Compute the difference in COP between the two models.
  6. Create a COP bubble map of the difference.
  7. Overlay the COP bubble map on the 2D oil thickness difference map.
Oil thickness difference with COP bubble map overlay for two user-selected reservoir simulation models within an ensemble.

Figure 2. Oil thickness difference with COP bubble map overlay for two user-selected reservoir simulation models within an ensemble.

CoViz 4D’s compatibility with Python allowed the team to write efficient and unambiguous scripts which performed the ensemble attribute calculations, in addition to the oil thickness calculations. The Jupyter notebooks’ user interface enabled the team to choose what data was being analyzed without having to change variables in the calculations by hand, permitting a higher degree of selectivity in the comparisons. The result was a simple Python workflow that facilitated computing what the team required, in addition to producing visualized results as interactive CoViz 4D models embedded in the Jupyter notebook.

See the CoViz 4D page for more information.

FURTHER READING

Forward Seismic Modeling in a North Sea Reservoir

Improved, flexible and faster access to critical data is vital for fully informed decision making in today’s multidisciplinary asset development teams. It is also imperative to include the highest fidelity quantitative temporal data when building reservoir models to improve development plans and maximize returns on investment.

4D Seismic Assisted History Matching—Closing the Loop in Reservoir Simulation

The traditional technique of manually adjusting reservoir simulation models to achieve a history match is time consuming and problematic at best; adding 4D seismic data to the equation only increases the complexity of the task. The inclusion of 4D seismic, however, results in better constrained simulation models and a greater understanding of the uncertainty inherently associated with models.

Oil Reservoir Compaction; Risk Reduction for Drilling Infill Wells

During production, North Sea chalk reservoirs are commonly prone to severe compaction due to reduction in pore pressure—this increases the load on the rock matrix, resulting in compaction. In some extreme cases the result is seabed subsidence on the order of 10 meters. This magnitude of subsidence causes challenges to the existing production infrastructure, including the integrity of the drilled wells, and leads to geomechanical changes in the subsurface which can impact seismic travel times.

Reservoir Simulation Ensemble Analysis via CoViz 4D Visualization and Jupyter Notebook Calculations

Dynamic Graphics provides powerful, interactive functionality to simplify software usage, but it has always recognized the need to prevent the limitations of a “black box” environment, allowing for the extension and customization of the tools by providing users with access to the underlying modular functionality. This enables clients to build their own solutions for precise problem solving, analysis, and 4D visualization without having to navigate the existing buttons and toolbars.

FOLLOW US

Accessibility Tools
hide