Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

Home > Theory and Equations > Deconvolution > Deconvolution

Deconvolution
Deconvolution is a mathematical tool that extracts the drawdown typecurve from the rate and pressure history.
Essentially the deconvolution process consists of the following steps:

Generate a typecurve as an initial guess

Superpose this typecurve with historical rate data to calculate synthetic pressures
Calculate the error between the calculated pressures and the measured pressures
Generate a new typecurve and repeat the process until the error between calculated and measured pressures is
minimized

The deconvolved typecurve can then be analyzed using conventional diagnostic analysis techniques to determine
various reservoir characterisitcs such as permeability ( ), skin ( ), reservoir size, well location and more.
Deconvolution provides an alternative to conventional diagnostic analysis and can show additional flow regime
information that would not normally be seen within the specified time frame of the buildup test. Please note that
deconvolution is a purely mathematical process and should be used with caution. The following guidelines show the
ideal conditions for deconvolving data to obtain the best possible results:

Data does not contain a lot of noise

Data is free from outliers in both rate and pressure
Rate history is reliable
Buildup typecurves used for deconvolution are consistent with each other
Wellbore and reservoir properties do not change significantly with time
A good estimate of initial pressure ( ) is available

For more information on how Deconvolution works, please refer to the following links to the technical videos.
Deconvolution in Well Testing, Part 1: Classical Methods
Deconvolution in Well Testing, Part 2: Modern Methods

Creating a Deconvolution Case

To create a deconvolution case go to the Deconvolution tab and click on the Create Deconvolution Case button
located within the Deconvolution Manager tab. A new tab will be created using the current data from the Production
Editor.

Interface
The diagram below shows the standard deconvolution interface. Note that any of the plots can be docked or undocked
by double clicking anywhere on the plot area.

1 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

Deconvolution Plots
Total Test

Displays the pressure and rate history. Pressure points can be selected (light blue) or de-selected (white) for fitting on
this plot.

Deconvolved Typecurve

Displays the deconvolved typecurve (unit rate pressure response) for the reservoir Note that until the deconvolution is
run this plot only shows the initial typecurve specified in the Task Bar and does not show the final result.

Data Typecurve

Displays the match between the deconvolved and selected typecurve data. Additional flow periods can be displayed
(overlayed) on this plot one of two ways:

1. By toggling the Show TC check boxes on the Task Bar.

2. By right-clicking on the Total Test plot over a flow period and selecting Show Typecurve from the pop-up
menu.

Task Bar

2 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

The Task Bar provides an easy way of displaying flow periods, selecting points, and selecting the initial typecurve.
Each row corresponds to either a drawdown (DD) or a buildup (BU) period. The Task Bar can be docked, undocked,
or hidden to allow for more display area when needed.

Show TC - Toggle these checkboxes to show/hide any flow period on the Data Typecurve plot.

Select Periods - Toggle these checkboxes to select/deselect any flow period pressure data for fitting.

Initial TC

Click on a radio button to select which flow period typecurve to use as the initial Deconvolved
Typecurve. By default the final buildup is selected. The initial typecurve can also be manually adjusted
as follows:

1. By clicking and dragging any derivative point (red) on the Deconvolved Typecurve plot.
2. Using the Linear Point Alignment option from the toolbar on the Deconvolved Typecurve plot,
click on a derivative point (red) and draw a trend line that the derivative points should follow.
When the left mouse button is released the derivative points will be aligned along this trend line.

Please note, after any change to the derivative the Pu curve will be updated accordingly.

Control Panel
The Control Panel allows modification of various deconvolution parameters/options and displays the fitting results.
Through this panel the deconvolution fit can be started and stopped.

3 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

Control Panel Toolbar

Filter The Data- Opens the filtering dialog (Data Reduction Module) used to filter out noisy/unreliable
data. Refer to the Filtering section for more information.

InitializeTypecurve- Resets the Deconvolved Typecurve to the selected initial typecurve (Init TC) on
the Task Bar.

Run Deconvolution - Starts the fitting (deconvolution / minimization) process to determine the
Deconvolved Typecurve that best fits the data.

4 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

Stop Deconvolution - Stops the fitting (deconvolution / minimization) process if it is running.

Anchor / Initial Pressure

Click on a radio button to select which pressure to anchor to in order to determine the initial pressure (pi) used for
fitting. The three options are:

pi - Use specified pi (no anchor); note that the specified pi can be entered in the cell or set by clicking
and dragging the calculated pressure curve on the Total Test plot up or down. Toggling the check box on
for this parameter will include it during the fitting process. It is recommended that when pi is not known
that the other anchoring options be used to minimize instability during the fitting process.

pwfo - Anchor to the final flowing pressure before the last buildup selected, pi is back calculated so the
calculated pressure curve on the Total Test plot matches this point.

Any Point

Anchor on any pressure point. After selecting this option click on the pressure point on the Total Test
plot to anchor to. A # symbol appears over the pressure point you are anchored to when this option is
on. To change the anchor point at a later time click on the anchor button on the Total Test plot toolbar
and click on another pressure point.

Typecurve Options
These parameters directly affect the Deconvolved Typecurve as well as the calculated pressure.

NDer - Sets the number of points on the Deconvolved Typecurve derivative. Increasing this value will
add more resolution to the typecurve but will reduce the calculation performance during fitting.

Pu1 - Sets the value of the first point of the Pu curve on the Deconvolved Typecurve. Toggling the
check box on for this parameter will include it in the fitting process. Note that the fitting process may take
longer to run when this parameter is set to auto. We also do not recommend fitting on this parameter
and pi on at the same time, to avoid instability in the fitting process.

Weighting Parameters
These parameters can be used to place more emphasis on various error terms to fine tune the fit. By default all the
weighting factors are set to 1.0 (no weighting). For more information on these parameters and how they affect the fit
please refer to the Theory section.

Wpres - Sets the weighting factor that the pressure error (Epres) is multiplied by during the fit.

Wcurv - Sets the weighting factor that the curvature error (Ecurv) is multiplied by during the fit.

Wrate - Sets the weighting factor that the rate error (Erate) is multiplied by during the fit. Note this option
is only visible when the Adjust Rates option is checked.

Adjust Rates - Toggle this check box on to include rate corrections in the fit. If this check box is not

5 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

checked, the Wrate and Erate parameters are hidden.

Deconvolution Results
Displays the following calculated errors during the fit. For more information on these parameters please refer to the
Theory section.

Epres - The overall error between the measured and calculated pressures.

Ecurv - The curvature error. Represents the amount of curvature on the derivative of the Deconvolved
Typecurve. The more linear the derivative curve the smaller this value will be.

Erate - The overall error between the measured and calculated rates.

ETLS - The total least squares error which is a summation of all the errors.

Advanced Options
Minimization Method

Sets the method to use for fitting (minimization). The default recommended method is More-Hebdon.
Other methods (Line Search and Double Dog Leg) can be used in this order if the current method fails
to find a solution.

Initialization Method

Sets the method used to initialize the Deconvolved Typecurve. The default method is Use Initial TC
which will initialize the typecurve to the flow period specified in the Initial TC column of the Task Bar.
The Radial method initializes the typecurve as pure radial flow (straight line) based on the flow period
specified in the Initial TC column of the Task Bar.

Filtering
Filtering is used to remove noisy and/or unreliable data (e.g. outliers) and reduce the number of points to improve
performance when fitting. For these reasons we recommend filtering the data prior to setting up or running the fit.

There are two ways to filter data:

Option 1

Click on the Filter The Data button on the Control Panel toolbar to open the filtering dialog (Data
Reduction Module). Modify the filtering options as needed to filter the data accordingly. Any filter
settings that are changed will be saved and used again in subsequent filter operations.

Please note that each time you use this filter option all the flow periods (Select Periods column on the
Task Bar) will be re-selected in order to include them in the filtering process.

6 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

Option 2

Select the Filter Data menu option from the pop-up menu on the Total Test plot. The filter will be applied
only to the flow period the mouse was over when right-clicking. Note that this filter option does not
remove any points from the dataset but rather de-selects the pressure points based on the filter so they
are not included when fitting. The filter applied is defined in the filtering dialog discussed in Option 1.

Theory
Assuming the reservoir and wellbore properties do not change over time, the governing equation being linear with
respect to pressure, and using the principle of superposition the following equation can be derived:

Where Pu is the unit rate pressure response of the reservoir.

If we define:

The superposition equation can be re-written as:

Where is the variable used in the least squares minimization process to calculate the derivative and Pu such that the
total least squares error ETLS is minimized.

The error terms used to determine the minimum error are defined as follows:

Where pm is the measured pressure and pc is the pressure calculated from the preceding superposition equation.

7 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

Where qm is the measured rate and qc is calculated by the least squares minimizer. Erate is zero unless the Adjust
Rates option has been turned on..

The curvature error (Ecurv), is defined as the amount of curvature on the derivative curve. it is a measure of how
smooth the derivative is, so that the smoother the curve the smaller Ecurv. For more detailed equations on this
term please refer to SPE papers 71574 and 77688.

The total least squares error (ETLS) is defined as follows:

Where wx is the weighting parameter for each Ex error term. These values can be adjusted to improve the fit when
needed by placing more emphasis on specific error terms.

Tips to Achieve Better Results

If you are not satisfied with the results try the following:

1. Filter your data to remove outliers and reduce the number of points.
2. Use as few flow periods as possible and assess whether or not the typecurve selected as the initial typecurve
(Init TC) is a good initial guess.
3. When selecting more flow periods it is important that they are as consistent as possible. To determine this,
select each flow period one at a time and compare their Deconvolved Typecurves.
4. If you have good quality measured point(s) in a de-selected flow period, try selecting these points individually.
5. Manually adjust the derivative on the Deconvolved Typecurve to achieve a better initial guess.
6. The overall results can be drastically affected by pi, especially with high rates. If you have a more reliable value
for pi we recommend using it instead of the defaults provided.
7. Avoid autoing on either pi or Pu1, especially at the same time, as this can significantly reduce performance and
produce poor results.
8. Try turning the Adjust Rates option on.
9. For noisy data increase the curvature weighting (wcurv) to make it easier for the algorithm to match the data.

FAQs
Why doesn’t the typecurve for a certain flow period show up on the Data Typecurve plot?

1. Check to see that the Show TC check box on the Task Bar has been checked for the flow period.
2. The range is too short, or has very few measured pressure points in it, and thus the typecuve cannot be calculated.

How do I select a different flow period to be the initial guess for the Deconvolved Typecurve?

Go to the Task Bar and select the appropriate flow period using the radio buttons in the Init TC column or go to the
Total Test plot, right-click on the desired flow period, then select Initial Typecurve from the pop-up menu.

8 of 9 5/13/2016 9:36 PM
Deconvolution http://www.fekete.com/SAN/TheoryAndEquations/WellTestTheoryEqua...

References
1. This deconvolution method is based on the work of Thomas Von Schroeter et al. and Michael M. Levitan et al.
for more information on the concepts of deconvolution please refer to the following papers: SPE 71574, SPE
77688, SPE 84290, SPE 90680.

Copyright © 2012 Fekete Associates Inc.

9 of 9 5/13/2016 9:36 PM