The goal of rOpenserver is providing an R interface to Petex (Petroleum Experts) applications Prosper, GAP and MBAL to perform automated tasks, generate datasets for statistical analysis, and advanced fine control of solvers and calculations.
rOpenserver is not in CRAN yet but in the meantime, you can install it from GtiHub using the devtools package, from the rOpenserver repository with:
install.packages("devtools") devtools::install_github("og-analytics/rOpenserver", dependencies = TRUE)
The argument dependencies has the role of downloading and installing packages that are key for rOpenserver, such as R6 and RDCOMClient. If you want to install RDCOMClient separately you could install it from its repository via devtools as well. Then after you could proceed to install rOpenServer.
devtools::install_github("omegahat/RDCOMClient")
In R, as in Python or any other programming languages, there are some conventions to deal with the filenames and folder or directories. In Windows, we should use a double-backslash to separate the paths. In Unix operating systems, like Linux and MacOS, we use a single slash.
For instance, in the following example, we set the folder where typically the Prosper examples are. Then , we use slash / or the double-backslash \\ to separate the folder names, otherwise you will get error. We use both ways in the following two code blocks.
# point to the well model prosper_folder = "C:/Program Files (x86)/Petroleum Experts/IPM 10/Samples/prosper" model_file <- "T00_Integrated_Oil_Well.Out" model_filename <- file.path(prosper_folder, model_file) file.exists(model_filename) #> [1] TRUE
This short example will read some values from the Reservoir pressure and temperature from a model residing in the Examples folder: T00_Integrated_Oil_Well.Out.
# load OpenServer library(rOpenserver) # Initialize OpenServer prosper_server <- openserver() # Start Prosper cmd = "PROSPER.START" prosper_server$DoCmd(cmd) #> [1] 0 # point to the well model prosper_folder = system.file("models", package = "rOpenserver") model_file <- "T00_Integrated_Oil_Well.Out" model_filename <- file.path(prosper_folder, model_file) file.exists(model_filename) #> [1] TRUE # open Prosper model open_cmd <- "PROSPER.OPENFILE" open_cmd <- paste0(open_cmd, "('", model_filename, "')") DoCmd(prosper_server, open_cmd) #> [1] 0 # get reservoir pressure cmd <- "PROSPER.SIN.IPR.Single.Pres" print(prosper_server$DoGet(cmd)) #> [1] "5200.000000000" # get reservoir temperature cmd <- "PROSPER.SIN.IPR.Single.Tres" print(DoGet(prosper_server, cmd)) # S3 class: another way of getting values #> [1] "210.000000000" # shutdown Prosper Sys.sleep(3) command = "PROSPER.SHUTDOWN" prosper_server$DoCmd(command) #> [1] 0 prosper_server <- NULL
rOpenserver
The previous example just shows a spark of the functionality of rOpenserver: starting Prosper, opening a well model, read values from it, and close Prosper. This is what we call an instance.
There are plenty more examples that make use of rOpenserver. During my years as Production Engineer working on well models and optimization we practically had endless use for this interface. Okay, you could call it a data science oriented solution, but not all is really data science. The use cases go beyond that. Let me cite a few, just for Prosper:
Batch automation for reading all the Prosper well models residing in your disk, or the network drive
Convert all the well models in Prosper to a rectangular or tidy dataset, very much like an Excel spreadsheet: a dataframe, which is a more powerful data structure for performing data analysis
Perform an inventory of all the fluid properties correlations that work in your field
Read specific variables inside the well models and presenting them in dataframes
Perform statistical analysis of missing values and outliers across all the variables that were entered in the well model
Treat the well models as observations (rows) instead of binary files, where the well properties are the variables or features.
Extract well variables for a particular well system, such as well tests, bottomhole data, correlation curves, pressure and temperature at depth, gas lift properties, or any other artificial lift method
Perform corrections of well data in batch, always keeping in mind that each well is a row, not a file
Export the well model from binary to CSV or another format of your preference. Excel, maybe.
Find and diagnose quicker a well model that refuses to run
Perform the calibration of the well model in an automated fashion
Connect the well model using a unique identifier to the master production database of the company
Use R as a common platform to connect the well models to SQL queries of the company production database
Reduce drastically the time of calibrating all the well model with new well test data, from months to weeks, or even days
Run multiple instances of Prosper to analyze wells that have multiple strings, such as dual gas wells: one Prosper instance for the Short string, and a second instance for the Long string