CA3217811A1 - Method and system for predicting sand failure in a hydrocarbon production well and method and system for producing hydrocarbon fluids from an earth formation - Google Patents

Abstract

Bottom hole pressure (BHP) actuals and bottom hole temperature (BHT) actuals in a hydrocarbon production well are recorded as a function of time, during production of hydrocarbon fluids. Selective statistical measures of both the BHP actuals and the BHT actuals are determined as a function of time. These selective statistical measures suitably represent estimates of expected normal BHP actuals and BHT actuals in case there is no imminent sand failure, supplemented with an uncertainty measure of the estimates of expected normal BHP and BHT. The BHP actuals and the BHT actuals are compared with respective BHP and BHT anomaly thresholds based on the selective statistical measures. An anomaly alert is automatically issued upon meeting a condition wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold. The anomaly alert is an indication of a predicted imminent sand failure of the hydrocarbon production well in operation.

Description

METHOD AND SYSTEM FOR PREDICTING SAND FAILURE IN A
HYDROCARBON PRODUCTION WELL AND METHOD AND SYSTEM FOR
PRODUCING HYDROCARBON FLUIDS FROM AN EARTH FORMATION
CROSS REFERENCE TO RELATED APPLICATIONS
The present application claims priority benefits of US Provisional Application No.
63/185,446 filed 7 May 2021.
FIELD OF THE INVENTION
The present invention relates to computer-implemented method and a computerized system for predicting sand failure of a hydrocarbon production well in operation. The invention further relates to a method of producing hydrocarbon fluids from an Earth formation, whereby applying the method and/or system for predicting sand failure, and a system for producing hydrocarbon fluids from an Earth formation comprising the system for predicting sand failure.
BACKGROUND TO THE INVENTION
Sand production has been a major concern to the oil and gas industry for decades.
Sand production is a consequence of fluid flow into a wellbore from a sanding-prone reservoir. Many well completions have been provided with sand control means, such as gravel packs and screens, in an effort to keep the sand inside the formation.
However, such sand control means are vulnerable and can ultimately lead to integrity failure due to sand ingress.
Various sand monitoring strategies have been proposed in the past, a number of which are discussed in SPE paper 189213 (P.Y. Lee and S. Frederiksen Kasper, "The 7 Sins of Managing Acoustic Sand Monitoring Systems," 2017). In this paper, results of a full field risk-based analysis is described. Wells with intact sand control in were considered likely to have either failed sand packs in the annular space or as a minimum to have sand failure occurring in the open hole. While sand control mechanisms are in place to prevent sand from migrating into the wellbore and topside facilities, sand migration still occurs.
The paper describes various means of sand detection methods that are available to mitigate sand migration through prudent well operation, while maximizing oil production. These methods were ranked, Acoustic Sand Detector method ranked as the primary option for sand monitoring. Significant issues were encountered, nonetheless.
SUMMARY OF THE INVENTION
In accordance to one aspect of the present invention, there is provided a computer-implemented method of predicting sand failure of a hydrocarbon production well in steady operation, comprising:
- recording bottom hole pressure (BHP) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- determining selective statistical measures of both the BHP actuals and the BHT
actuals as a function of time;
- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds; and - automatically issuing an anomaly alert upon meeting a condition wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold.
In accordance with another aspect of the invention, there is provided a computerized system for predicting sand failure of a hydrocarbon production well in steady operation, comprising:
- a memory for recording bottom hole pressure (BHP) actuals in a hydrocarbon production well as a function of time, during stable production of hydrocarbon fluids, for recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during stable production of hydrocarbon fluids;
- a processing unit configured to execute computer readable instructions for:
- determining selective statistical measures of both the BHP actuals and the BHT actuals as a function of time;
- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds;
and

- 2 -- automatically issuing an anomaly alert upon meeting a condition wherein both the BHP
actuals and the BHT actuals exceed their respective anomaly threshold; and - an output device for outputting said anomaly alert.
In another aspect of the invention, there is provided a system for producing hydrocarbon fluids from an Earth formation, comprising:
- a hydrocarbon production well penetrating into a hydrocarbon fluid containing region of an Earth formation; and - the computerized system for predicting sand failure of the hydrocarbon production well in operation as defined above.
In still another aspect of the invention, there is provided a method of producing hydrocarbon fluids from an Earth formation with a hydrocarbon production well, comprising:
- allowing hydrocarbon fluid from the Earth formation to flow into the hydrocarbon production well in a bottom hole location;
- recording bottom hole pressure actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during (stable) production of hydrocarbon fluids;
- determining selective statistical measures of both the BHP actuals and the BHT
actuals as a function of time;
- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds; and - automatically issuing an anomaly alert upon meeting a condition wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold.
BRIEF DESCRIPTION OF THE DRAWINGS
The drawing figures depict one or more implementations in accordance with the present teachings, by way of example only, not by way of limitation. In the figures, like reference numerals refer to the same or similar elements.
Fig. 1 schematically shows a hydrocarbon production well comprising a system for predicting sand failure;

- 3 -Fig. 2 schematically shows a block diagram illustrating a method for predicting sand failure in a hydrocarbon production well;
Fig. 3 schematically shows a bock diagram illustrating a specific implementation of the method illustrated in Fig. 2;
Fig. 4A shows a graph of actuals, rolling averages, and anomaly thresholds as a function of time for BHT according to one example;
Fig. 4B shows a graph of actuals, rolling averages, and anomaly thresholds as a function of time for BTP of the same example as Fig. 4A;
Fig. 5A shows a graph of actuals and anomaly thresholds as a function of time for BHT according to another example;
Fig. 5B shows a graph of actuals and anomaly thresholds as a function of time for BTP of the same example as Fig. 5A;
Fig. 5C shows a graph with data from an acoustic sand detector of the same example as Fig. SA;
Fig. 5D shows a graph of choke valve position of the same example as Fig. 5A;
and Fig. 5E shows a graph of a production stability indicator in the same example as Fig. 5A.
DETAILED DESCRIPTION OF THE INVENTION
The person skilled in the art will readily understand that, while the detailed description of the invention will be illustrated making reference to one or more embodiments, each having specific combinations of features and measures, many of those features and measures can be equally or similarly applied independently in other embodiments or combinations.
Bottom hole pressure (BHP) actuals and bottom hole temperature (BHT) actuals in a hydrocarbon production well are recorded as a function of time, during production of hydrocarbon fluids. It has been found that by continuously monitoring these actuals, particularly during stable production conditions, an imminent sand-induced well impairment can be predicted before the sand has actually been produced to an extent of causing a said control failure. These actuals are continuously monitored by a computerized anomaly detection method, which issues an anomaly alert when certain condition is met in the BHP and BHT actuals that is indicative of high risk of sand control failure. This allows for adjusting a production parameter, upon the anomaly alert being issued, in order to

- 4 -mitigate production of sand before a well sand control failure incident has materialized.
Although not the only way, but an effective way to mitigate the production of sand is by reducing drawdown pressure.
The computerized anomaly detection method herein includes determining selective statistical measures of both the BHP actuals and the BHT actuals, as a function of time.
These statistical measures should be defined by the BHP and BHT actuals which are determined contemporaneously during well-stable operating conditions. The selective statistical measures may, for example, represent estimates of expected normal BHP actuals and BHT actuals in case there is no imminent sand control failure risk or sever well impairment, supplemented with an uncertainty measure of the expected normal BHP and BHT actuals. Based on the selective statistical measures, BHP and BHT anomaly thresholds are derived. The BHP actuals and the BHT actuals are compared with the respective BHP and BHT anomaly thresholds, and an anomaly alert is automatically issued upon meeting a condition wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold. The anomaly alert is an indication of a predicted imminent sand control failure and/or severe impairment of the hydrocarbon production well in operation.
It has been found that by basing an anomaly alert on the requirement that both the BHP and BHT actuals exceed anomaly thresholds, a high precision in predicting sand failure can be achieved (by avoiding false positives) while at the same time a high recall accuracy can be achieved (by avoiding false negatives).
The anomaly detection methodology is governed by a number of anomaly detection parameters, which include statistical analysis parameters that control how the selective statistical measures are determined. These parameters can be selected (by modelling and/or empirically) to optimize the discriminating performance for sand failure incident predictions. The parameters may even be adjusted on the fly, to further optimize precision and recall as described above. In addition, by adjusting these anomaly detection parameters and/or the statistical analysis parameters the same anomaly alert principles may be used for detecting other causes for anomalies such as early well impairment conditions or issues which are not related specifically to sand failure predictions. Several versions of the same anomaly detection loops may be run at the same time, in order to provide other anomaly alerts in addition to the anomaly alert that is triggered by high risk of imminent sand failure.

- 5 -In one example, the anomaly detection may employ decision rules, such as Western Electric rules, Nelson rules, or Westgard rules, to determine whether an anomaly threshold has been exceeded. However, the invention is not limited to any one particular selection of rules.
Drawdown pressure is a term of art which indicates a differential pressure that drives fluids from the Earth formation into the wellbore. The drawdown pressure of a producing interval may typically be controlled by surface or subsea chokes. The presently proposed methods and systems may be used to empirically determine the maximum drawdown pressure that may be safely applied during production before damage or unwanted sand production occurs.
In the context of the present disclosure, the term Bottom Hole Temperature (BHT), also known as downhole temperature, is the temperature of the hydrocarbon production well measured at a point of interest in the well in a vicinity of the producing Earth formation during the hydrocarbon production process. The term Bottom Hole Pressure (BHP) is the pressure in the hydrocarbon production well measured at the point of interest in the well in said vicinity of the producing Earth formation during the hydrocarbon production process. The BHP and BHT are measured on the wellb ore side of any sand control means, i.e. at the low-pressure side of the drawdown (the formation-facing side of the sand control means being the high-pressure side).
Furthermore, in this description references will be made to the terms true positive, false positive, true negative, false negative. True positive (TP) means an anomaly alert is triggered, and an impairment or sand failure was indeed imminent or present.
False positive (FP) means an anomaly alert is triggered while there is no impairment or imminent sand failure. False negative (FN) means that no anomaly alert is triggered, despite .. presence of impairment or imminent sand failure. True negative (TN) means no anomaly alert is correctly triggered as indeed there is no impairment or imminent sand failure. False negatives should preferably be avoided as much as possible, as these could lead to severe impairment and possible well damage.
Finally, reference is made to derived indicators precision and recall.
Precision is the percentage of true positives of all positives, i.e. precision = TP/(TP+FP) x 100%. It is a measure of nuisance created by the anomaly detection analysis (low precision corresponding to high nuisance). Recall is the percentage of impairment or sand failure events that are correctly detected, i.e. recall = TP/(TP+FN) x 100%. Recall is a measure of

- 6 -effectiveness of the anomaly detection analysis, as success is defined as avoiding false negatives.
Fig. 1 schematically shows a hydrocarbon production well 1 comprising a system for predicting sand failure of the hydrocarbon production well in operation. The hydrocarbon production well can be an open hole well, such as shown, or a perforated-cased well (not shown). In either case, the well may be completed with a sand control device 8, such as a sand screen 4 and/or a gravel pack 5 in the annulus 6.
The hydrocarbon production well 1 penetrates a hydrocarbon fluid containing region 11 of an Earth formation. Hydrocarbon fluids 10 can flow from the Earth formation into the well in a bottom hole location. Generally, the bottom hole location is considered to be below a production packer 3. Also in the bottom hole location, below the production packer 3, there is provided a BHP and BHT sensor 2 within the hydrocarbon production well. The hydrocarbon fluids 10 typically flow to surface 12 via a production tubing. These sensors may be combined in one unit or they me be provided as separate gauges.
Signals representing BHP actuals and BHT actuals are transmitted to surface 12, possibly via a gauge cable 14 or any alternative route. For instance, the BTP and BHT sensors may be provided in the form of fiber optic distributed sensing or bragg grating sensing, in which case the signals are transmitted optically.
At surface, there is provided a computerized system 20 for predicting sand failure of the hydrocarbon production well 1, in operation. This system may comprise an input interface 21 connectable to the BHP and BHT sensors 2. The interface 21 may suitably be part of, or integrated with, a distributed control system (DCS) used to operate the hydrocarbon production well 1. Such DCS routinely gathers readings or measurements from all sensors, and operates final control elements such as chokes and shutdown valves.
Signals representing BHP actuals and BHT actuals pass through the interface 21 to a computer memory 23. The computer memory 23 suitably comprise a time series database and/or be serviced by a Data Historian, such as for example an OSIsoft PI
System.
The computerized system 20 further comprises a processing unit 25, configured to execute computer readable instructions for determining selective statistical measures of both the BHP actuals and the BHT actuals, as a function of time. These selective statistical measures suitably may represent rolling estimates of expected normal BHP
actuals and BHT actuals, in case there is no imminent sand failure, and an uncertainty measure of these estimates. In other words, these statistical measures may be used to "extrapolate" the

- 7 -"normal" BHP and BHT progression associated with stable operating conditions, and provide a statistical uncertainty of the extrapolations.
The selective statistical measures may be determined based on BHP actuals within a predetermined first rolling time window, and on BHT actuals within a predetermined second rolling time window. The rolling time windows can be set such as to optimize the best rolling estimates. The duration of the rolling time windows may be selected, and it is also an option to employ weighted rolling time windows which attribute higher weights to the actuals at certain times within the rolling time windows relative to other times within the rolling time windows. Furthermore, the parameter(s) that define the first predetermined rolling time window, used on the BHP actuals, may be set at the same or different value(s) as those of the second predetermined rolling time window, which is used on the BHT
actuals.
In the examples described below, the selective statistical measures are specifically based on determining rolling averages of the BHP and BHT actuals within their respective rolling time windows, as well as rolling standard deviations on these rolling averages, all as a function of time. However, any other selective statistical measures that describe the evolution of the BHP and BHT actuals corresponding to normal stable operating conditions may be employed, including for example parameter fitting to an appropriate mathematical function or model.
The first and second predetermined rolling time windows should be selected such that gradual normal time progressions in BHP and BHT actuals are closely flowed by the rolling averages, but that faster abnormalities in BHP and BHT actuals, which may be associated with well impairment events, are allowed to deviate from the rolling averages.
For typical hydrocarbon production wells, the predetermined rolling time windows may be set at about 10 days or longer, preferably 20 days or longer. This allows for detecting anomalies that occur on the time scale of 1 or several days, or faster. The minimum values determine the accuracy by which the normal drift behavior is "extrapolated" in case of an anomaly occurring faster than normal. The maximum values for the predetermined rolling time windows may be 90 days or shorter, preferably 60 days or shorter, more preferably 40 days or shorter. The maximum value determines how well the normal drift of BHP
and BHT actuals is followed by the rolling averages. In particular studies, Applicants have found 30 days to be a suitable rolling time window for the purpose of detecting imminent sand failures.

- 8 -The processing unit 25 further is programmed to detect relevant anomalies of the BHP and BHT actuals from their normal trends. For the purpose of the present objectives, an anomaly is relevant if it is associated with well impairment or an imminent sand control incident or failure. The processing unit 25 may be programmed to automatically issue an anomaly alert signal in case an anomaly is detected. The computerized system 20 may further comprise an output device 27. The output device 27 may be in communication with the processing unit 25 (either directly or indirectly via, for example, the DCS) for outputting an anomaly alert upon receiving the anomaly alert signal from the processing unit 25.
The processing unit 25 may comprise one or more microprocessors or central processing units (CPU). Although in Fig. 1 the processing unit 25 is represented as a single unit, in practical implementations, its functionality, or part thereof, may suitably be integrated with the Data Historian or PI system, or otherwise distributed over multiple processors.
When an anomaly alert is issued, a production parameter of the hydrocarbon production well may be adjusted, to (pre-emptively or reactively) mitigate or prevent the production of sand, leading to a sand control failure. The adjusting of the production parameter may for instance be aimed at effectively reducing the drawdown pressure. This may be accomplished, for instance, by choking back or otherwise reducing the rate of hydrocarbon production. In a practical implementation, an operator 28 may receive the anomaly alert, and determine whether a mitigation action is justified and the operator will determine the course of action to mitigate the higher risk of imminent sand control failure.
The operator 28 may intervene via the DC S, for example to regulate a production choke 7.
This intervention may be an initial intervention, which may at least delay a sand control failure event. During the initial intervention, a decision may be made whether other corrective measures to address the well impairment situation may have to be taken as well, such as a well stimulation operation. An ultimate determination of the most appropriate corrective action may involve further evaluation of well performance data. It is not excluded that the role of the operator 28 may in future be assisted by an advisory algorithm, or partially or fully be taken over by an advanced controller. The advanced controller may involve several machine learning algorithms to be able to translate all relevant well performance data to appropriate and effective control actions.

- 9 -In one group of embodiments, the processing unit 25 may be programmed to define anomaly thresholds over time for both the BHP and the BHT. The anomaly thresholds are suitably based on the selective statistical measures, and they are used to detect instances where the BHP and BHT actuals significantly differ from the rolling estimates.
Significant deviations from the estimates are indicative that the hydrocarbon production well is operating in a state of anomaly. In the examples below, the anomaly thresholds may respectively be based on the BHP rolling average and the BHP rolling standard deviation, and on the BHT rolling average and the BHT rolling standard deviation.
For the purpose of detecting higher risks of imminent sand control failures, it has been found that on hydrocarbon production wells that produce liquid-dominant fluids, the BHT threshold should be set on the high side of the BHT rolling average or other BHT
rolling estimate. Conversely, on gas-dominant producing wells, the BHT
threshold should be set on the low side of the BHT rolling average or other BHT rolling estimate. The BHP
anomaly threshold is usually on the low side of the BHP rolling average or other BHP
rolling estimate, regardless of which type (liquid-dominant or gas-dominant) of hydrocarbon production well.
The anomaly thresholds may suitably be further determined by control limit factors.
The control limit factors for the BHP and BHT may be selected equal to each other or they may be individually be predetermined. The control limit factors determine by how many standard deviations the BHP and BHT actuals are allowed to deviate from their respective rolling estimates (in one direction) before an anomality warning is triggered.
If the control limit factors are set too low, then the system is prone to issuing false positives as small normally occurring deviations will be interpreted as an anomaly. However, the higher the control limit factors are set, the larger a deviation must be to be considered an anomaly.
The system would then be more prone to missing true events leading to more false negatives. In particular studies, Applicants have found a control limit factor of 2.0 to be suitable for both the BHT as well as the BHP. It is expected general that any value between about 1.5 and 3.0 could be found to yield satisfactory results.
In some cases, there can be strong glitches in either one of the BHT and BHP
actuals, which should preferably not be interpreted as positive indicators for imminent sand failure events. One way to avoid such glitches to trigger an anomaly alert, is to apply a predetermined alert delay time. This alert delay time signifies during how much time the BHP actuals and the BHT actuals must exceed their respective anomaly thresholds before

- 10 -an anomaly alert can be triggered. By selecting a longer alert delay time, the precision of the anomaly detection (i.e. distinguishing true positives from false positives) increases but with the risk of paying a price in effectiveness of the analysis (recall). The reason is that the rolling estimates will slowly creep towards the actuals, causing the actuals to drop back to below the anomaly thresholds as a result of the analysis method (a not because the root cause for the anomaly signals has been removed). Generally, to safeguard effectiveness, the event delay times should therefore be a small percentage of the rolling time window lengths. Applicants have found an event delay of 10% or lower of the rolling average time window length to be suitable, preferably 5% or lower. Lower is better, if the user can live with the lower precision (i.e. greater relative number of false positive alerts).
An implementation of the computer-implemented method for predicting sand failure of a hydrocarbon production well in operation can be outlined as Fig. 2. At 100, BHP and BTH actuals from the hydrocarbon production well are recorded. At 110, the selective statistical measures are determined from the BHP actuals and the BHT actuals, to provide rolling estimates of BHP and BHT during stable production conditions, and indications of statistical uncertainties therein. The statistical measures are determined as function of time, and they may suitably be determined over the actuals within predetermined rolling time windows. At 130, BHP and BHT anomaly thresholds over time are determined.
These are based on the respective rolling averages and rolling standard deviations. At 140, the BHP
actuals and the BHT actuals with the respective anomaly thresholds. Only when a condition is met, at 150, wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold, then an anomaly alert is automatically issued at 160.
Fig. 3 illustrates a specific implementation where the statistical measures as determined in 110 are specified as determining BHP and BHT rolling averages of the BHP
and BHT actuals at 115 and determining BHP and BHT rolling standard deviations of the respective rolling averages at 120.
Figs. 4A and 4B illustrate an example of what the implementation of Fig. 3 may look like on a liquid-dominant producing well. Fig. 4A shows BHT actuals 31, BHT
rolling average 32 (over 30-day time window), and BHT anomaly threshold 33, all as a function of time during stable operation of the well. The BHT anomaly threshold 33 is computed by adding 2.0xBHT rolling standard deviation to the BHT rolling average 32. Fig.
4B shows BHP actuals 41, BHP rolling average 42 (over 30-day time window), and BHP
anomaly threshold 43, all as a function of time (contemporaneous with panel A). The BHP anomaly

-11-threshold 43 is computed by subtracting 2.0xBHP rolling standard deviation from the BHP
rolling average 42. At time ti, the BHT actual 31 equals the BHT anomaly threshold 33 and exceeds the BHT anomaly threshold 33 thereafter. At time t2, the BHP
actual 41 equals the BHP anomaly threshold 43 and "exceeds" the BHT anomaly threshold 33 thereafter. Exceeding, in this context, means that the actuals are further removed from the rolling average than the anomaly threshold. The moment t2 that both BHP and BHT
actuals exceed their respective anomaly threshold, a pre-determined alert delay time td is triggered and if both BHP and BHT actuals continue to exceed their respective anomaly threshold during td then the anomaly alert is automatically triggered at t3 =
t2 + td. The anomaly alert may stay active until t4, when one of the BHP actuals 31 or the BHT actuals 41 no longer exceeds the respective anomaly threshold.
The effectiveness of the described method and system has been assessed by retroactively applying the method/system to wells that have been known to succumb to sand control failure. An example is shown in Figs. 5A to E. All five graphs show data that is contemporaneous in time. Fig. 5A shows BHT actuals 31 and BHT anomaly threshold 33, both as a function of time. Fig. 5B shows BHP actuals 41 and BHP anomaly threshold 43, both as a function of time. Fig. 5C shows a signal from an acoustic sand detector (ASD) that was installed in the production flow line downstream of the production choke 7. Fig. 5D shows the choke 7 position, expressed in a percentage of opening, and Fig. 5E
shows when the production was stable. Stable production is defined as at least 6 hours of time have elapsed after the last change in choke position (or, said differently, a constant choke position for at least the past 6 hours).
Some of the BHT actuals 31 and BHP actuals 41 have been ignored, for instance when the production was not stable or when there were clear outliers (spikes).
It can be seen at time ti the BHT actuals 31 cross the BHT anomaly threshold 33 and at t2 the BHP actuals 41 cross the BHP anomaly threshold 43. Depending on the predetermined alert delay time td the system could have issued an automatic anomaly alert at time t3 when the amount of actual produced sand was still below the detection limit of the ASD. However, the method had not been applied to this well, and at time ts a catastrophic sand control failure occurred as seen by the sudden and dramatic increase in the ASD signal. Immediately following this event, the choke 7 was closed. In this case, since the sand control failure has already become fact, the well had to stay shut in for years. However, this could have been avoidable by partially choking back shortly after t3.

- 12 -A control limit factor of 2.0 has been applied for both BHT and BHP, and the alert delay time was set at 1 day. Had a higher control limit factor been used, then the BHT and BHP actuals would have crossed their respective anomaly thresholds later, but this could still have been on time to avoid the catastrophic failure. Applicant has varied parameters .. of the method and applied the method retroactively to a large number of wells. Lowering the alert delay time from 3 days to 1 day, resulted in an improvement of recall from 87% to 93%, at the cost of precision dropping from 61% to 41%. Further optimization of the parameters can be empirically made to further improve recall, and preferably also precision.
The method and system described herein functions best during stable production of the well. This can be accomplished by ignoring BHT and BHP actuals at times of not-stable production, and not using the actuals at these times when determining the selective statistical measures. Stable production may be defined as production with a constant production choke position whereby the choke position has not been subject to change within the last 2 hours, preferably within the last 5 hours.
The person skilled in the art will understand that the present invention can be carried out in many various ways without departing from the scope of the appended claims.

- 13 -

Claims (20)

1. A computer-implemented method of predicting sand failure of a hydrocarbon production well in operation, comprising:
- recording bottom hole pressure (BHP) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- determining selective statistical measures of both the BHP actuals and the BHT
actuals as a function of time;
- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds; and - automatically issuing an anomaly alert upon meeting a condition wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold.

2. The computer-implemented method of claim 1, further comprising a predetermined alert delay time, wherein the condition comprises that both the BHP actuals and the BHT
actuals exceed their respective anomaly threshold for a duration of at least the alert delay time.

3. The computer-implemented method of claim 1, wherein recording of said BHP
actuals comprise sensing a BHP with a pressure sensor configured within the hydrocarbon production well; and wherein recording of said BHT actuals comprise sensing a BHT with a temperature sensor configured within the hydrocarbon production well.

4. The computer-implemented method of claim 1, wherein the hydrocarbon production well comprises completion equipped with a sand control device.

5. The computer-implemented method of claim 4, wherein the sand control device comprises one or both of a sand screen and a gravel pack.

6. The computer-implemented method of claim 1, wherein said selective statistical measures represent estimates of expected normal BHP actuals and BHT actuals in case there is no imminent sand failure, and an uncertainty measure of the estimates.

7. The computer-implemented method of claim 6, wherein said selective statistical measures are determined based on BHP actuals within a predetermined first rolling time window and on BHT actuals within a predetermined second rolling time window.

8. The computer-implemented method of claim 7, wherein each uncertainty measure corresponds to a standard deviation of an average; and wherein determining said selective statistical measures comprises:
- determining a BHP rolling average of the BHP actuals, and a BHP rolling standard deviation, both as a function of time, and both based on the BHP actuals in the predetermined first rolling time window;
- determining a BHT rolling average of the BHT actuals, and a BHT rolling standard deviation, both as a function of time, and both based on the BHT actuals in the predetermined second rolling time window;
and wherein said BHP anomaly threshold over time is based on the BHP rolling average and the BHP rolling standard deviation and said BHT anomaly threshold over time is based on the BHT rolling average and the BHT rolling standard deviation.

9. The computer-implemented method of claim 8, further comprising a predetermined BHP control limit factor which determines how many times the BHP rolling standard deviation the BHP anomaly threshold is removed from the BHP rolling average and at each time and/or further comprising a predetermined BHT control limit factor which determines how many times the BHT rolling standard deviation the BHT anomaly threshold is removed from the BHT rolling average and at each time.

10. The computer-implemented method of claim 1, wherein adjusting a hydrocarbon production parameter, upon an anomaly alert being issued, in order to mitigate production of sand.

11. A computerized system for predicting sand failure of a hydrocarbon production well in operation, comprising:
- a memory for recording bottom hole pressure (BHP) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids, for recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- a processing unit configured to execute computer readable instructions for:
- determining selective statistical measures of both the BHP actuals and the BHT actuals as a function of time;

- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds;
and - automatically issuing an anomaly alert upon meeting a condition wherein both the BHP
actuals and the BHT actuals exceed their respective anomaly threshold; and - an output device for outputting said anomaly alert.

12. The computerized system of claim 11, further comprising an input interface connectable to a pressure sensor configured within the hydrocarbon production well and to a temperature sensor configured within the hydrocarbon production well whereby signals representing BHP actuals and BHT actuals pass through the interface to the memory.

13. The computerized system of claim 11, wherein said determining of selective statistical measures of both the BHP actuals and the BHT actuals as a function of time comprises:
- determining a BHP rolling average of the BHP actuals, and a BHP rolling standard deviation, both as a function of time, and both based on the BHP actuals within a predetermined first rolling time window;
- determining a BHT rolling average of the BHT actuals, and a BHT rolling standard deviation, both as a function of time, and both based on the BHT actuals within a predetermined second rolling time window;
and wherein said BHP anomaly threshold over time is based on the BHP rolling average and the BHP rolling standard deviation and said BHT anomaly threshold over time is based on the BHT rolling average and the BHT rolling standard deviation.

14. A method of producing hydrocarbon fluids from an Earth formation with a hydrocarbon production well, comprising:
- allowing hydrocarbon fluid from the Earth formation to flow into the hydrocarbon production well in a bottom hole location;
- recording bottom hole pressure (BHP) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- determining selective statistical measures of both the BHP actuals and the BHT
actuals as a function of time;

- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds; and - automatically issuing an anomaly alert upon meeting a condition wherein both the BHP actuals and the BHT actuals exceed their respective anomaly threshold.

15. The method of claim 14, further comprising a predetermined alert delay time, wherein the condition comprises that both the BHP actuals and the BHT actuals exceed their respective anomaly threshold for a duration of at least the alert delay time.

16. The method of claim 14, wherein recording of said BHP actuals comprise sensing a BHP with a pressure sensor configured within the hydrocarbon production well;
and wherein recording of said BHT actuals comprise sensing a BHT with a temperature sensor configured within the hydrocarbon production well.

17. The method of claim 14, wherein said selective statistical measures represent estimates of expected normal BHP actuals and BHT actuals, both in case there is no imminent sand failure, and an uncertainty measure of the estimates.

18. The method of claim 14, wherein said determining of selective statistical measures of both the BHP actuals and the BHT actuals as a function of time comprises:
- determining a BHP rolling average of the BHP actuals, and a BHP rolling standard deviation, both as a function of time, and both based on the BHP actuals in a predetermined first rolling time window;
- determining a BHT rolling average of the BHT actuals, and a BHT rolling standard deviation, both as a function of time, and both based on the BHT actuals in a predetermined second rolling time window;
and wherein said BHP anomaly threshold over time is based on the BHP rolling average and the BHP rolling standard deviation and said BHT anomaly threshold over time is based on the BHT rolling average and the BHT rolling standard deviation.

19. The method of claim 14, wherein adjusting a hydrocarbon production parameter, upon an anomaly alert being issued, in order to mitigate production of sand.

20. A system for producing hydrocarbon fluids from an Earth formation, comprising:
- a hydrocarbon production well penetrating into a hydrocarbon fluid containing region of an Earth formation; and - a computerized system for predicting sand failure of the hydrocarbon production well in operation, wherein said computerized system comprises:
- a memory for recording bottom hole pressure (BHP) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids, for recording bottom hole temperature (BHT) actuals in a hydrocarbon production well as a function of time, during production of hydrocarbon fluids;
- a processing unit configured to execute computer readable instructions for:
- determining selective statistical measures of both the BHP actuals and the BHT actuals as a function of time;
- defining a BHP anomaly threshold and a BHT anomaly threshold over time, based on the selective statistical measures;
- comparing the BHP actuals and the BHT actuals with the respective anomaly thresholds;
and - automatically issuing an anomaly alert upon meeting a condition wherein both the BHP
actuals and the BHT actuals exceed their respective anomaly threshold; and - an output device for outputting said anomaly alert.