Automatic DCA (aDCA)
Overview
This user task is designed to calculate an Arps decline curve fit to either oil or gas production for a well. The task can take as input one or more well properties as well events which determine when a decline curve starts. A new forecast will be performed when a new well event occurs, as well as when a user-specified amount of new production data is added to the well. Optionally, the task can automatically calculate change points where the decline curve starts. If there are no well events and the automatic change point calculation is not selected, the user task will use the date corresponding to the maximum rate as the initial point for DCA.
The decline curve forecast is either done “deterministically”, with an optimization procedure for best fit, or probabilistically, using the Nested Sampling procedure. The decline curves are displayed in a plot shown in the “Plots” folder found under the user task node for the well. Either the current deterministic or probabilistic fit is displayed in the plot, depending on the user selection.
In addition to the rate forecast, the user task outputs several well properties if K-A is configured to store them (see Prerequisites, below). Specifically, the task stores for each forecast the Di and b from the Arps equation, the EUR for the oil or gas production, and the cumulative oil or gas production at the end of the forecast period. If the forecast is deterministic, a single property value is output. If the forecast is probabilistic, multiple property values are displayed (P1, P10, P50, P90, and P99 percentiles).
Prerequisites
Well events: The AutoDCA user task can take as input one or more well properties which indicate dates of workovers, choke changes, and other events where a new decline curve should be fitted to the wells. Before creating the AutoDCA user task, the well events must first be specified in the global list of well properties. The global list of well properties can be seen under the “Automation” view, in the “Well properties” tab (shown in the figure below). These well events can be specific to the company and must of type “Event”. Well properties are identified by a unique alias, shown in the first column of the well property table below.
 |
When defining the AutoDCA task in K-A, you must set up the well event input in the Parameters tab as shown in the figure below. In the figure below, there is a single input of Type “Well property” which defines the well events, entitled “Well events” (highlighted). Note however that more than one input for well events is allowed. The Property dropdown box in the UI (highlighted) corresponds to the “Name” of the well event (the second column in the table above). Note that the property name is what is visible in the property table and is used only for display purposes, hence configurable.
 |
Well properties The Auto DCA user task also outputs well properties which are optional but recommended to take full advantage of the user task. These well properties are described in the table below.
Well property | Alias | Measure | Description |
|---|---|---|---|
Average Arps Di decline | DCA_Average_Arps_Di_Decline | Inverse time | Arps Di decline rate value from best fit, or weighted arithmetic average of Di from sampling if performing probabilistic forecast |
Percentile Arps b decline | DCA_{percentile}_Arps_Di_Decline | Inverse time | Replace the tag {percentile} by P1, P10, P50, P90, and P99. This property represents the percentile of the Arps Di decline rate calculated from sampling if performing a probabilistic forecast. |
Percentile Arps b exponent | DCA_{percentile}_Arps_ b_Exponent | No unit | Replace the tag {percentile} by P1, P10, P50, P90, and P99. This property represents the percentile of the Arps b exponent calculated from sampling if performing a probabilistic forecast. |
Percentile Cumulative Phase Forecast | DCA_Average_Cumulative_{phase}_Production | Gas or liquid volume | Replace the tag {phase} by either oil or gas. This property represents the cumulative production up to the end time of the forecast, including the historical production. The value is calculated from the best fit curve, or from a weighted arithmetic average from the sampling. |
PAverage EUR Cumulative Phase | DCA_ Average_EUR_Cumul_{phase} | Gas or liquid volume | Replace the tag {phase} by either oil or gas. This property represents the cumulative EUR, including the historical production. The EUR is calculated from the best fit curve, or from a weighted arithmetic average from the sampling. The input value to “EUR abandonment rate” is used as the final rate to calculate the EUR. If this value is not specified, then the default value is 0.01 times the maximum historical rate. |
Average Cumulative Phase Prod Forecast | DCA_{percentile} _Cumulative_{phase}_Production | Gas or liquid volume | Replace the tag {percentile} by P1, P10, P50, P90, and P99. This property represents the percentile value of the cumulative production up to the end time of the forecast, including the historical production, as determined from the sampling procedure. Consequently, the values are calculated only when a probabilistic forecast is performed. |
Percentile EUR Cumulative Phase | DCA_ {percentile}_EUR_Cumul_{phase} | Gas or liquid volume | Replace the tag {percentile} by P1, P10, P50, P90, and P99. This property represents the probabilistic cumulative EUR, including the historical production. The input value to “EUR abandonment rate” is used as the final rate to calculate the EUR. If this value is not specified, then the default value is 0.01 times the maximum historical rate. The values are calculated only when a probabilistic forecast is performed. |
Average Arps b exponent | DCA_Average_Arps_b_Exponent | No unit | Arps b exponent from best fit, or weighted arithmetic average of b from sampling if performing probabilistic forecast |
Input
This section describes the input to run the AutoDCA user task. The description of each user task input is provided in the table below
Perform DCA on gas - Boolean indicated whether gas or oil is the main phase.
Rate - The dataset which contains the rate data (gas or oil).
(One or more well events) - As described in the “Prerequisites” section above, the user task can be configured to accept one or more well properties of type “Event”, indicating when significant well events occurred to start a new decline curve analysis. This can be empty.
Automatically select change points - Boolean to indicate whether the user task should automatically detect and use change points to determine when to start a decline curve analysis.
Window Size - Only showing if “Automatically select change points” is true. Controls the amount of detail in the analysis of the rate curve. In general, the smaller the window size, the more change points will be detected.
Filter peak size - A non-negative number indicating the number of consecutive points to treat as peaks or valleys. If this value is > 0, the user task will attempt to filter out extreme peaks (valleys) in the rate curve to remove potentially anomalous rate values.
Lag time for forecasting - The amount of time that the user task will wait until performing another DCA calculation.
Forecast duration - Duration of forecast beyond the last data point recorded.
Perform probabilistic forecast - If true, the user task will perform a probabilistic forecast. Otherwise, a one-curve, best fit will be performed.
Max DCA calls - Only visible when “Perform probabilistic forecast” is true. Often unnecessary to modify unless the forecast does not appear to converge correctly, or if the calculations are taking too long.
Error in rate (fraction) - Only visible when “Perform probabilistic forecast” is true. Controls the error tolerance around the rate data.
Output
The user task outputs several datasets, described in the table below.
Decline points - This dataset indicates the location of the change points, which are either derived from the well event input(s), and/or the automatic selection.
Filtered rates - The filtered rates which occur when “Filter peak size” is positive.
P1/P10/P50/P90/P99 - The output of the decline curve fit. When performing deterministic forecasts, only the P50 output will be employed and will represent the best fit. When performing a probabilistic forecast, all percentiles will contain output and are a reflection of the curves and the input error fraction.