Workflows for history matching using enkf
- Preparing an ECLIPSE reservoir model for use with ERT.
- Creating an observation file for use with ERT.
- Creating a configuration file for ERT.
- Prior distributions available in ERT.
- Customizing the simulation workflow in ERT.
ERT works with the ECLIPSE family of reservoir simulators, and can be used for sensitivity and uncertainty analysis as well as to condition reservoir models to dynamic data using the EnKF algorithm.
This document describes a step-wise workflow for using ERT for uncertainty analysis and history matching. It covers every step from starting the ERT executable to exporting the final results. However, it is not a complete description of ECLIPSE, and it is assumed that the user:
- Has basic ECLIPSE knowledge and a stable ECLIPSE reservoir model.
- Has basic UNIX knowledge. E.g., can move around the file system and edit plain text files.
Step 0: Ensuring LSF access (optional)
When available, the ERT can use LSF to execute simulation jobs on a cluster. Using LSF is generally preferred, as it speeds up the clock-time of the workflow considerably. However, direct LSF access is not available everywhere.
This step is entirely optional, but if you choose to run ERT from a workstation without direct LSF access, you will be restricted to the local and rsh queue systems.
Step 1: Locating the executable
The first step in the workflow is to ensure that you are able to execute the ERT binary. The ERT prototype is a Linux only application, and depending on where you compiled it it should be available as /path/to/ert/devel/libenkf/applications/ert_tui/ert.
ERT can be started from the command line by running the binary ert from the command line with a configuration file as argument.
$ /path/to/ert/devel/libenkf/applications/ert_tui/ert config_file
where config_file is created according to Creating a configuration file for ERT.
Step 2: Preparing your ECLIPSE model
Unlike other assisted history matching tools ERT relies heavily on the restart capability of ECLIPSE. Thus, some preparatory steps must be taken to make your reservoir model ready for use with the ERT. In essence, you need to modify your ECLIPSE data file so that:
- An ECLIPSE simulation of the model can be started from any point in the file system.
- A non-unified restart file and summary file are written at each report step (i.e. at each DATES, TIME and TSTEP keyword in the ECLIPSE schedule file.).
- The layout of the active grid blocks is invariant between realizations.
The necessary modifications are explained in Preparing an ECLIPSE reservoir model for use with ERT.
Step 3: Defining observations and associated uncertainties
In the EnKF algorithm not only the observed values, but also their associated uncertainty play a key role. Thus, when using the ERT for conditioning to dynamic data, you need to provide an explicit definition of which observations to condition on and an associated uncertainty for each observation. These definitions are specified in a separate observation file, see Creating an observation file for use with ERT.
Although providing an observation file is not required when using the enkf application for sensitivity and uncertainty analysis, it is strongly recommended, as observed values will not appear in plots without it.
Step 4: Creating a configuration file
To tailor the Ensemble Reservoir Tool (ERT) to your specific problem, you need to create a configuration file. The configuration file serves several purposes, which are:
- Defining which ECLIPSE model to use, i.e. giving a data, grid and schedule file.
- Defining which observation file to use.
- Defining how to run simulations.
- Defining where to store the results.
- Creating a parametrization of the ECLIPSE model.
Creation of a configuration file is described in Creating a configuration file for ERT.
Step 5: Using your parametrized ECLIPSE reservoir model
There is no definite answer on how to proceed from here. In fact, it very much depends on the problems you are adressing. However, we have tried to provide a set of jobs in Using a parametrized ECLIPSE model with ERT which can serve as building blocks for a workflow tailored to your problem.