This tool provides a method to generate semi-realistic synthetic reservoir data to support reservoir simulation efforts, spatial modeling, and machine learning workflows. In its current state, the program is able to generate permeability, porosity, depth, and thickness values for any x-y-z combination of grid block geometries. These values are then exported to your working directory as .txt files which can be used in any external modeling program or can be read back into python for further use.
This project was created to support Dr. Matthew Balhoff at the University of Texas at Austin in the creation of a reservoir simulation textbook. This project would not be possible if not for Dr. Michael Pyrcz's geostatistical python library, "GeostatsPy" which provides the technical backbone through which the simulated data is created. Additionally, portions of the code that create the porosity and permeability values were borrowed from Dr. Michael Pyrcz's workflow titled "GeostatsPy_synthetic_well_maker".
- Install GeostatsPy in your anaconda environment by writing 'pip install geostatspy' in your anaconda terminal.
- Download the GSLIB executable files (here) and move them into your working directory.
- Fill out the input template excel file and make sure the excel file is also in the working directory.
- Run the program in Jupyter Notebook.
- Check to see if the export files were sucessfully created in the working directory.
To begin, you must have GeostatsPy installed on your anaconda environment. To do this, simply open up a terminal within anaconda and type 'pip install geostatspy'.
After that, you must also have the GSLIB executable files downloaded within your working directory. To download the executable files, go to GSLIB.com and navigate to the download page. For Mac users this may be tricky as you have to jump through some extra hoops. Here is a link to further intstructions for installing the necessary executables for Mac users. Once you have the executable files downloaded, make sure sgsim.exe is in your working directory or else the program will not work properly.
To use the program, an excel file must be created that contains the various inputs of the program. Within the GitHub repository is a template of the input sheet or you can download it directly here.
The inputs are as follows:
nx ny nz [int]represent the number of grid blocks in the corresponding x, y, and z directions.hsiz [float]represents the horizontal dimensions of each grid block. I.e the length of each grid block in the x and y directions.vsiz [float]represents the vertical height of each grid block. I.e the height of the grid block in the z direction.por_mean & por_std [float]represent the mean and standard deviation of the target porosity distribution.perm_mean & perm_std [float]represent the mean and standard deviation of the target permeability distribution.DP [float [0:1]]represents the Dykstra-Parsons coefficient of the total permeability distribution across all layers. A measure of permeability heterogeniety.hor_ang [float]represents the angle by which the reservoir is rotated about the z-axis (out of the page) according to the right hand rule. Think azimuth.ver_ang [float]represents the angle by which the reservoir is rotated about the y-axis (north). Think dip.
The program can be run either in a Jupyter notebook or from a command terminal. Note, interactive plotting is only currently supported within the Jupyter notebook version.
To run the program, simply instantiate the object and then call its master() method.
SyntheticReservoir = SRS()
SyntheticReservoir.master()
Running the above code will import all of the necessary inputs from the excel file, generate the synthetic data, and then save the files to your working directory. If you have multiple rows within your excel file corresponding to different reservoir geometries (i.e nx-ny-nz combinations), the program will generate a unique export file for each row automatically.
Within the program there are three main visualization tools to let the user inspect various properties of the simulation.
self.plot_histogram()self.plot_layers()self.visualize()
The first method, self.plot_histogram() takes two inputs, the data used in the histogram and the desired x-label. The data used in the histogram is either self.por or self.perm.
The second method, self.plot_layers() takes three inputs: the data used to plot, the number of layers to visualize, and the plot title.
The third method, self.visualize(), takes the desired data (e.g self.por) as its only input. This method will generate an interactive 3D plot through %matplotlib qt. Note, when dealing with large reservoir geometries, this method may take a while to load.
If you encounter any problems with the program or have any general questions about its use feel free to email me at travis.salomaki@gmail.com.
Travis Salomaki, B.S. Petroleum Engineering Honors, the University of Texas at Austin. LinkedIn