Type Well

$\label{type-well}$

“Type Wells” are often referred to as “Type Curves”.
“Type Curves” refer to idealized production plots (based on equations and/or numerical simulation) to which actual production results are compared.
“Type Wells” are based on actual well production data and represent a typical production profile for a collection of wells for a specified duration.

What is a "typical" well?

The arithmetic average of a group of wells are commonly applied to estimate a type well. That said, experts (e.g. David Fulford) would argue that a geometric mean better approximates the median. Production rates and EURs are generally distributed lognormal, accordingly averaging the logs of the associated values make a lot of intuitive sense. We have option for both in whitson⁺.

1. Create a Type Well Scenario

$\label{add-type-well}$

Click Type Well in the navigation panel (just under Auto-Forecast).
To the upper right, click ADD TYPE WELL.
Provide a name for the scenario and select all the wells by clicking the check box at the top of the well list (left of Well Name) or just some of them by clicking the checkbox next to each well.
Click SAVE to the lower right.

You can store multiple type wells per project

This page gives you an overview of the created type wells, when they were created, who created them and what wells went into creating them.

2. Generating a Type Curve

2.1 Well Selection

$\label{TCwellselection}$

Click the checkbox next to search wells to include all the wells available in this type well scenario.

There are two ways to include or exclude individual wells from the Type Well creation:

Click the check-mark next to the well name.
CTRL + click time-series: exclude that well from type well.

You can also select the well by simply clicking the time-series datapoint on the graph. That will filter to only that well in the well menu to the left. An alternative is to search for the well name in the "Search Wells" input field.

2.2 Pick Relevant Phase

First, you must pick the phase you would like to build a type well for.

$\label{tc_pickingphase}$

Additionally, you can go to the Settings tab to set the x axis and time resolution, adjust the normalization parameter for the y axis and choose the normalization multiplier, switch between the averaging methods, add a previously saved autoforecast or single-well DCA to the wells, etc. as outlined in sections below.

3. Overview of Plot Options

Watch the video above to get an overview of the different plot options. This is the quick summary of the icons available to the top-right of the plot:

Show/Hide Well Count: Use the stairs icon to show/hide the number of wells versus time on the plot.
Show/Hide Individual well production data: Click the eye icon to enable/disable the plotting of individual well production data (in gray).
Show/Hide Legend: The legend also lets you activate P90, P50, P10 (Switched off by default).
Downloads: Download the type well rates, including the individual well rates displayed on the plot into an Excel sheet or export the plot as an image.
Other typical zoom options.

3.1. Plot Colors

$\label{changingcolors}$

Use the color palette icon to spread colors, color by attribute (such as well, data, reservoir properties or completion metrics) or switch back to default colors as shown in the GIF below.

Pro-tip

You can also copy and paste colors from one box to another - this trick is applicable in Comparison Plots and all other places where you are allowed to adjust colors manually in whitson⁺

3.2. Nice-to-know items

Additionally, in the well selection pane on the left, you have options to:

Start at peak time: Type curve is constructed from the peak rate, i.e. moves the peak rate of each well to start at time = 0.
Start all plots from time=0: Starts type curve construction by aligning the time=0 for each well.

Time Alignment

Time Align on First Production
* Strength: on larger well sets, communicates the production profile considering time to peak.
* Weakness: may not accurately reflect production decline behavior.

Time Align on Peak Rate Date
* Strength: more accurately reflects production behavior.
* Weakness: excludes ramp up time which might have a small impact on EUR but is important to first year revenue projection.

Search and Filter wells: The filtering is applied across the software, so you can filter elsewhere and find the same applied here and vice versa.
The X icon clears the search bar.
You can also use the arrow keys next to the well name to shift the production data of that well, to the left or right in time.

$\label{add-type-well}$

Highlighting individual wells across all plots

You can click the well name, in the well list. This highlights the well in all the plots (type curve plot, map and probit plot) in black for comparisons with the type curve, identifying the well location on the map, and identifying the relative ranking of the well in the Probit plot. This is shown in the image below -

$\label{highlightinall}$

Pro-tip

If you click the time series data point on the graph, it automatically searches the Well list to show the relevant well. This makes it easy to identify the well you are clicking. If you click the well name, this well is highlighted in black and you can clear the search bar by clicking the X icon to see all your wells again!

3.3 Probit Plot

$\label{9-probit}$

Represent the statistical distribution of something (e.g. EUR, IP60, physical parameter) at a point in time.
The shape can help to determine if the results trend towards a lognormal distribution.
A “probit best fit” regression can yield statistical insights including a measure of uncertainty (e.g. P10/P90 ratio)

4. Probabilistics

$\label{TCprobabilistics}$

The P10, P50 and P90 time series are not shown by default. Add them by clicking the legend.

5. Settings

5.1 Normalization

You should pick relevant ways to normalize the type well - restructures data to improve comparability.

x-axis: pick between normalized time, normalized flowing time (remove days with no production) or cumulative production.

There are 4 different x-axis time options available

Normalized Time: aligns time to first production.
Normalized Flowing Time (by total production): considers only time steps with active oil, gas, or water production, excluding periods with no production (all rates must be 0).
Normalized Flowing Time (by stream): considers only time steps when the rate of the chosen stream (oil, gas, or water) is non-zero, excluding periods with zero flow in the selected stream.
Cumulative Production: plots the cumulative production on the x-axis.

$\label{5-x-axis}$

y-axis: Used for dimensional normalization to pick between rate or rate normalized by properties such as lateral length, proppant pumped, fluid pumped, etc. This puts the wells into a meaningful comparative context. You can also use the normalization multiplier here to convert the normalized rates to likely actual rate for a presumed value for the normalization variable (ex. 10,000 ft lateral length)

$\label{6-norm}$

Time resolution: You can pick between monthly or daily.

Pro-tip

If you prefer not to autoforecast, if you save individual well decline curve forecast with the same name, it acts like a saved autoforecast case and selecting that name in this dropdown adds the forecasts for all the wells.

5.2 Averaging Method

$\label{7-avg}$

There are to options for calculating the average:

The geometric mean represents the central tendency of a set of values using the product of all values, with the nth root taken (where n is the number of values).
The arithmetic mean is the average of a set of values, calculated by summing up all values and dividing by the total count.

Geometric Mean

“The geometric mean better approximates the median. Medians are more predictive for the outcome of a small number of samples (in this case, drilled wells), as well as being much more robust to outliers. Since production rates and EURs are generally distributed lognormal, an average of the logs of values makes a lot of intuitive sense.“
- David Fulford

6. DCA - Forecast the Average vs Average the Forecasts

$\label{8-avg-forecasts}$

You can either Average the Forecast or Forecast the Average

What's the difference

Average the Forecasts

Time consuming without auto forecast option
Useful for statistical evaluation and P10/P90 quantification of EUR

Use the DCA Forecast dropdown in the Settings tab to do this.

Forecast the Average

Apply a decline to the truncated type well to obtain a full life profile of EUR
Time effective, but does not provide distribution of EURs

Use the DCA tab in the Type well to do this.

In this section we'll forecast the type well (average), the P90, P50, P10 curves using using DCA. Here, you can also force the DCA fit to honor the EUR of the type curve by toggling on the switch 'Set EUR_{type well} = EUR_{dca fit}'. You can save this DCA fit to the type curve (or P90, P50 and P10 curves) and toggle this on in the Settings tab to see this saved DCA fit in the main type well plot as shown in the GIF below.

$\label{TCDeclines}$

7. P10, P50 and P90: Percentile Lines vs Probit Plot

$\label{percentiles}$

7.1 Probit plot

Calculates percentiles based on the number of data points (wells) in a probit plot. It's only for one point in time.

7.2 Percentile curves

Calculates percentiles based on a vector of data points and a given percentile (p) using linear interpolation

7.3 Which approach is correct?

Both approaches are correct, but they serve different purposes and are suited for different types of data.

Probit Plot Percentiles: This approach calculates percentiles based on the total number of data points (wells) in a probit plot. It is suitable for situations where you have a fixed number of data points and you want to calculate percentiles relative to the total count. This method ensures that the percentiles are evenly spaced across the dataset, which can be useful for certain types of analyses, such as in probit plots.

Linear Interpolation Percentiles: This approach calculates percentiles based on the position within a sorted vector of data points. It is suitable for situations where you have a continuous distribution of data points and you want to interpolate between them to estimate percentiles. This method provides more flexibility and can handle situations where the number of data points may vary or where you need to estimate percentiles for specific positions within the dataset.

So, the "most correct" approach depends on the nature of your data and the specific requirements of your analysis. If you have a fixed number of data points and you're working with a probit plot, the first approach is more appropriate. If you're working with a continuous distribution of data points and you need to interpolate percentiles, the second approach is more suitable.

7.4 Why isn't the cumulative P50 exactly the same as the EUR P50 in the probit plot?

The P50 (also true for P10 and P90) value for each time step isn't linked to one actual well. On the contrary, it will (almost) always be a mix of multiple wells over time. That's fundamentally different than the P50 EUR (or P10 and P90), which is a result of one well only over time. In most cases, the values will be relatively similar.