History Matching

When the geological model is complete, you should verify that it reproduces the observed
rates and pressures during a simulation run. This is commonly used to prove that the
model makes a reasonable prediction for the future. Usually users run tens or maybe
hundreds of different simulations, in order to find the best match. The History Match
Analysis in Petrel allows you to easily and quickly analyze hundreds of simulation runs
to find the best match. It is recommended to start off at a high level and then examine in
more detail all the way down to obtain a data curve for every well.

The Calculations (History Matching)

The History match analysis tool calculates the absolute difference between the observed
and the simulated values at each time step and then divides by the number of points used,
to give users an averaged difference value or what we call the match value (M). It also
normalizes the value by dividing the difference by a normalization parameter σ:

Here M is the match value, N is the number of time samples, Si is the simulated values
and Oi the observed values at time i.

By setting the normalization values correctly, you can force the match value to be in a
given value range. Hence every vector will have the same importance in the combined
match. The normalization can also be used for weighting purposes, that is, choosing
weights of different size will put more weight on selected vectors in a combined match.

Zero values from the observed data can be filtered out. This way you can disregard data
from times when the well is shut in, so that if all values from the observed data are zero
for a well, the data points in this interval will not be used in the calculation.

History Matching Analysis

The History match analysis process is located under Simulation in the Processes pane.

If no History Match Statistics sets have been created then the first time the panel is
displayed users will only be able to select Create new match. A unique name for the
match set will be calculated automatically, but it can be edited via the appropriate edit
field in the panel. The Observed Data Set will contain a list of all the observed data sets
previously loaded and will default to the first one found.

If History Match Statistics sets have previously been generated then the default selected
option is Edit existing match. If a History Match Statistic set has been selected in the
Results Pane then it will be the default selection. The Observed Data Set shown will be
the one that was used to create the History Match Statistic set. If you do not want to edit
an existing History Match Statistic set, then simply select Create new match.

Select the Observed Data Set that you would like to use in this match set. In the white
box in the bottom part of the panel you can find all the vectors that have been found in
both the observed data sets and in all the cases. Set to Use Default for all, or highlight
any combination of vectors and deselect Use Default to specify individual settings. You
do not need to specify the wells or cases, as the History Match Analysis will be
automatically performed on all wells in all cases. The calculations are relatively fast so
this does not pose a problem.

 Note: If you change the Observed Data Set, the list of vectors will be recalculated and
any customized settings made to the old set will be lost.

Sample

Choose between Simulation frequency and Observed frequency.

Simulation frequency: In this instance, the observed data is sampled forward, that is, is
regarded to be constant from one observed data point to the next. The values for the
observed data are averaged and compared with the simulated data at every time step.
Observed frequency: In this instance, the simulated data is considered to be constant
from one time step and backwards to the previous time step. These values are compared
with the observed data at all observed data points.

Normalization

This process is used to normalize the values. More specifically:

1.  it will make the match unitless;

2.  it will make the result be in a certain ’normalized’ range;

3.  it can also be used as weighting when combining matches.

Choose between Average (%) and Absolute. The first option will divide the error
(difference between simulated and observed) with the given % of the average observed
value; the Absolute option will use the value you specify.

By using an Absolute value of 1, the match value will be the un-weighted average error
for the time interval in the original unit. The match value is an absolute value and will
always be positive, independent of whether or not the simulated value is less or more than
the observed.

Note: This stage of the process only performs the ‘expensive’ sampling and
normalization of the data. The remaining thresholding and summation of the partial
match values are done ’on the fly’ when toggling the various vector matches, identifiers
and cases, whilst displaying the plots/maps.

History Matching Statistics

The History Match Statistics folder is on the Results Tree and holds all the Match Sets. It
also contains two subfolders:

1.  Vector Matches contains all the different vectors (curves) that are available for
matching analysis;

2.  Identifier contains all the wells.

The combination of those and of what is selected in the Cases pane, will decide what is
calculated and displayed.

Settings

In the  Settings panel for a particular match set there is a Settings tab where you can
define threshold values.

Threshold
A threshold is used when displaying Qualitative matches. Qualitative matches are graded
from 1 (good) to 5 (poor). Any raw match value below ‘Min’ will be graded 1. All raw
values above ‘Max’ will be graded 5. The interval in between is divided into three
equally sized intervals, graded 2-4.

In cases where the user chooses to display Quantitative match values, the raw match
value will be displayed and the threshold value will not be used. The color map used for
Quantitative plots will be continuous, while for Qualitative plots it will be discrete.

Case Comparison

This is used when comparing two cases in map view. When you compare two cases it is
very unlikely that they will have exactly the same value down to the last decimal,
therefore, for practical purposes, we will consider two cases to be equal, where the
absolute difference is less than the Equality value set here.
Statistics

Under Settings for a particular Match set there is a Statistics tab. This contains the
information generated from running the initial part of the process (the sampling and
normalization of the data). This is a more detailed version of the data that is written out to
the log window when the History Match Analysis process is run. The table has six
columns:

Simulation & Vector

It displays the name of the case and the match vector.

Minimum & Maximum

It displays the minimum and maximum match values found. This should help you when
trying to determine the threshold settings.

Zero data

This reports whether zero data values have been found in both the observed and summary
data (this can happen if the well is shut in). Such information should help you determine
whether you need to switch on the Zero Data Filtering option in the main process panel.

Note: It is only possible to switch on or off the Zero Data Filtering on a vector-by-vector
basis. You cannot switch it on or off on a case-by-case basis.

Wells matched

This reports the number of wells that had both the observed and summary data for the
vector in the case. This is a simple way of checking that the expected number of wells
had data for the vector.
Style

Under Settings for History match statistics you find the Style tab, which has three further
tabs:

General

General style settings, such as Labeling, Marker size and color templates can be found
here. Optimizing the colors (min/max value range) when displaying your results in a map
window is recommended The Auto Size option automatically sizes the font used to
display the labels in such a way that they do not overlap. This can result in very small
labels that are difficult to read; therefore, the option is switched off by default. Printers
have a far higher resolution than a typical screen so it can be useful to switch this option
on, prior to producing a hard copy output.
Function window

Choose coloring to indicate the best and worst matches. Choose between single vector
mode and combined vector mode. Combined vector mode (see How combined vectors
are computed) will combine (sum) the selected matches when displayed. The combine
mode is also selectable via the toolbar and the popup menu of the History match
statistics set.

If the combine vector mode is selected then a single point will be plotted for all the
vectors using the Match color. If the single vector mode is selected then points will be
plotted for each vector selected. There are options on this panel for how the points should
be colored, either using the property from the subject or evenly across the current color
map.
Map window

Choose between Well compare mode and Case compare mode.

Well compare mode displays match values for wells for a selected case. Their value can
be displayed as either qualitative or quantitative. Bubble plots are available for this mode
where the size of the marker is determined by how good the match is. This makes it very
easy to see which wells are producing good and bad matches. Options are available in the
panel to specify the minimum and maximum marker size for the bubbles. Qualitative or
Quantitative matches can be selected using the radio buttons on this panel or via the
popup menu of the Case subject.

Case Compare mode compares two selected cases.

The aim of this example is to give an idea of how to conduct such an analysis and also
how to interpret the results you see.

Before you start

If you have not observed yet data and simulation results in your project, you will first
need to import some. Observed data can be imported onto the Wells folder in the Input
pane; ECLIPSE simulation results can be imported into the Case pane.

Run the history matching process

Configure the options as described above, under Settings for History match statistics,
then select Apply or OK. A new Match Set will be created under the History match
analysis folder in the Results pane.

General plotting notes

There is an option on the root menu of the Cases pane (right-click in the Cases pane) to
select or deselect all the cases. This option can be very useful as there is no top-level root
node to toggle on/off all child nodes.

There is no need to select a Match statistics set, as this will be done automatically once
you have selected a valid combination of vectors, wells and cases.

To make a vector or well display on the X-axis rather than on the Y-axis, right-click on
the appropriate vector or well, and choose Select as X.

Petrel will determine the type of plot you want, by analyzing the items you have
selected. It is therefore possible, with certain combinations, that a plot cannot be
generated. These scenarios will produce an empty plot; this should not be confused with
plots that are valid, but for which no valid data can be found. These are indicated by the
text “No Valid Data found” being written across the plot.

The wells and vectors are initially listed in the Results pane in the order that they were
found when the History match analysis process was run. It is possible to sort the wells or
vectors by name or property by right-clicking on the corresponding folder and selecting
one of the various Sort By… options.

The X- and Y-axis ranges will often change, due to different types of plots being

generated. To normalize the view.

Rank cases

Open a function window and then select your choice of:

2.  Vector Matches of interest (under the Vector Matches folder of the History match
statistics folder of the Results pane)

3.  Wells to include (under the Identifier folder of the History match statistics folder of
the Results pane)

The plot above shows Match Value Vs Case. This is a very easy way to find the good
and bad cases. The superimposed arrow indicates that the match value is decreasing when
we do more simulations. This shows that we are improving the model for this particular
match.

If a single vector is selected, its name will be displayed on the Y-axis. If multiple
vectors are selected, the name for the Y-axis will change to indicate that it is a
combination. If multiple vectors have been selected, then you can use the toolbar button
to display a marker for each vector match, rather than a single marker representing the
combined match for all.

Cross plotting vectors

Open a function window and then select your choice of:

1.  Cases of interest (from the Cases pane)

Wells to include (under the Identifier folder of the History match statistics folder of the
Results pane)3.

The plot above shows a cross-plot of Oil Production Rate Match Values Vs a
Combination of other Match Values. This can be used to see trends in your data. An
example would be if you were able to improve one match while something else is made
worse. This type of plot will very quickly show you any negative correlations. The box,
in the lower left corner, indicates the optimal position for the match values to appear.
Those are the cases with low difference in Match Values for all Vectors being plotted.
The superimposed green arrow indicates a good trend (both are improving) where as the
superimposed red arrow indicates a bad trend (one match value is improving whilst
another is getting worse).

Cross plotting wells

Open a function window and then select:

1.  Cases of interest (from the Cases pane)

2.  Vector Matches of interest (under the Vector matches folder of the History match
statistics folder of the Results pane)
3.  At least two wells (under the Identifier folder of the History match statistics folder
of the Results pane). One well must be designated as the X-axis. To do this, right-click
on it and select the menu option ‘Select as X’. The other wells will be combined, if
necessary, and placed on the Y-axis.

Plotting well matches in map view

Be sure to select Well compare (Single case select mode) in the Map window tab of the
Settings panel for History match statistics, or deselect the Case compare  toolbar
button.

Open a map window and then select:

1.  Case of interest (from the Cases pane)

2.  Vector Matches of interest (under the Vector matches folder of the History match
statistics folder of the Results pane)

3.  Wells to include (under the Identifier folder of the History match statistics folder of
the Results panel)
The map above shows the qualitative match for a specific case on a map window. Values
are given by coloring, and the bubble plot mode has been enabled so that the size reflects
the mismatch. We immediately see the three red wells that match poorly. Switch between
qualitative and quantitative in the History Match Statistics Settings or via the popup menu
options of the case.

Note: Only one case at a time can be selected in this mode. Nothing will be displayed
until only one case has been selected. Once the map is displayed, selecting a case will
automatically deselect the previous one.

Comparing two cases in map view

Be sure to select Case compare (dual case select mode) in the Settings panel for
History match statistics, or press the Case Compare toolbar button.

Open a map window and then select:

1.  The two cases of interest (from the Cases pane).

2.  Vector Matches of interest (under the Vector Matches folder of the History match
statistics folder of the Results pane)

3.  Wells to include (under the Identifier folder of the History match statistics folder of
the Results pane)
The coloring of wells will be green, yellow and red. Green indicates that the second case
is better than the first. Red indicates that the second is worse. Yellow means that the two
cases are equal (within the equality +/- range set in the settings dialog for the Match set).

Note: Only two cases at a time can be selected in this mode. Nothing will be displayed
until exactly two cases have been selected.  One case will have a displayed next to it and
the other case a. If you select a different case then it will become the new second case and
the will move to that case. If you deselect the case marked with the then the next case
selected will become the new first case and the will move to that case.

How combined vectors are computed

When selecting more than one well and/or vector type, Petrel calculates a merged history
matching value “on the fly”. The calculation is performed in the following way:

where the denominator is the number of match vectors added in the numerator.