Tuesday 7 June 2016

Experimental Design for Reservoir Simulation

ED4Sim v2p2

ED4Sim is a spreadsheet tool designed to assess uncertainty in petroleum reservoirs using reservoir simulation.  It is a very powerful spreadsheet, that has state of the art capabilities in an easy to use flexible package.




This spreadsheet was developed mainly while working at Nexen, and along with the RSMPlots spreadsheet, the company has allowed it to be released publicly. 

Update June 7th, 2016

The link to the Essential regression add-in is broken and I am unable to find this tool elsewhere. This is surprising as I consider it the best free regression tool and as good or better than commercial (not free) tools. I have provided links below to my copies of this tool.




Instructions for use of the spreadsheet, and contact information are contained within the instructions in the spreadsheet itself.


Quick Start Guide

Download and install the Essential Regression Excel add-in from my copies here:
64 bit version
32 bit version

The original link for Essential Regression is below, but is broken.
http://www.jowerner.homepage.t-online.de/download.htm

Download the ED4Sim spreadsheet from the download links below.  Open it and follow the instructions on the instruction sheet.  If you are asked if you want to update links when opening, select "Don't Update".


Downloads

Version 2.2 is the current version for Office 2010.  This version allows the RSM files fom Eclipse to have more than 65000 lines, which was a limitation in Excel 2003 and earlier.

Version 1.6 is the last version for Excel 2003 and earlier.


What it does

  • Allows input of up to 30 factors, and associated information.
  • Select from Central composite, Box Behnken, Plackett-Burman, space filling (Nearly Orthogonal Latin Hypercube) designs, OPAT(One Parameter At a Time) or Fractional Factorial designs.
  • Simulation data deck deck can be copied into workbook and uncertainty parameters set with "wildcards".
  • Exports simulation data files for ED runs.
  • Imports results from simulator result (RSM) files.
  • Prepares data for essential regression (ER).
  • Integrates with results from ER.
  • Uses regression results as response surfaces in Monte Carlo simulation.
  • Up to 10 results (response variables) can be analysed and run through MC.
  • Allows entry of Monte Carlo input distributions for each factor.
  • Runs the Monte Carlo (does not need any add-in Monte Carlo software).
  • Generates Pareto plots from regression results.
  • Creates Component plus Residual plots, accounting for interaction and square terms.
  • Generates Histograms of results and input uncertainties.
  • Generates Tornado charts from the response surface, accounting for interaction and square terms.
  • Helps select, generate and run selected realizations through the reservoir simulator.
  • Can select realizations (at P90, P50 or P10 level) using Monte Carlo, constrained by up to 10 results.
  • Exports realizations, and imports results to compare to predictions.

Acknowledgements

Space filling nearly orthogonal latin hypercube designs are from:
Sanchez, S. M.  2005.  NOLHdesigns spreadsheet. Available online at:
http://diana.cs.nps.navy.mil/SeedLab/

Regression models are created using Essential Regression, available at: http://www.jowerner.homepage.t-online.de/download.htm

Screenshots

List of steps and macro buttons, colours keep track of steps completed.



Factor input, note the factor type and number of levels.


Many designs are available and can be browsed by colour code. 


Data format is created for ER and application launched. 


Essential Regression is used for regression models.


Many distribution types are available for factors uncertainty, including users defined distributions.


Pareto charts are generated from Essential Regression results including standardized pareto charts.


Plots Factor distributions in both design values and real scale values. 


Creates statistics and histograms of results from Monte Carlo simulation. 


Creates Tornado plot from the response surface accounting for square and interaction terms.


Relative uncertainties pie chart from Tornado plot results.


Components plus residual plots for any response surface and all factors, accounting for square and interaction terms.