SYSTEM AND METHOD FOR ONLINE AUTOMATION

• US 20110220410A1
• Filed: 09/14/2009
• Published: 09/15/2011
• Est. Priority Date: 10/14/2008
• Status: Active Grant

• Current Assignee: Schlumberger Technology Corporation (SLB)
• Original Assignee: Schlumberger Technology Corporation (SLB)
• Inventors: Dunlop, Jonathan, Belaskie, James, Aldred, Walter
• Granted Patent: US 8,838,426 B2
• Time in Patent Office: Days
• US Class Current: 175/26
• CPC Class Codes: E21B 2200/22 Fuzzy logic, artificial intelligence, neural networks or the like
• CPC Class Codes: E21B 44/00 Automatic control systems specially adapted for drilling operations, i.e. self-operating systems which function to carry out or modify a drilling operation without intervention of a human operator, e.g. computer-controlled drilling systems; Systems specially adapted for monitoring a plurality of drilling variables or conditions
• CPC Class Codes: E21B 7/04 Directional drilling
• CPC Class Codes: G05B 13/04 involving the use of models or simulators

Abstract

A changepoint detector for modeling data received from at least one sensor in a proces in the hydrocarbon industry. The data is segmented into a plurality of segments and fo each segment a model is assigned and the data corresponding to the segment fit to tha model. A plurality of segmentations are thus provided and these segmentations ar evaluated and assigned weights indicative of the fit of the models of the segmentation t the underlying data. The segmentation models are further used to calculate a result tha may be input to a process control program.

Claims

1. A method for automating or partially automating a process in the hydrocarbon industry, where the hydrocarbon-industry-process is subject to a change in one or more operating conditions, can be controlled by at least one parameter and is monitored by at least one sensor providing an input data stream, comprising:

receiving the stream of input data from the at least one sensor;

upon receiving a new data item in the input data stream, using a processor to postulate that the input data stream is segmented according to a plurality of possible segmentations, wherein each of the plurality of possible segmentations comprises a plurality of segments divided by one or more changepoints, and wherein the changepoints are indicative of a change in at least one of the one or more operating conditions;

evaluating each of the plurality of possible segmentations by;

fitting portions of the input data corresponding to each segment in the segmentation being evaluated to a model corresponding to the each segment in the segmentation being evaluated; and

determining how well the models for the segments of the segmentation being evaluated fit the portions of the input data corresponding to each segment of the segmentation being evaluated;

generating an output from at least one of the plurality of possible segmentations and the models corresponding to the segments of the at least one of the plurality of possible segmentations; and

using the output to control at least one parameter of the hydrocarbon industry process.

2. The method of claim 1, further comprising:

periodically removing from consideration segmentations having evaluation results indicative of poor model fit of the models corresponding to the segments of the segmentation.

3. The method of claim 2, further comprising:

upon receiving additional data in the input data stream, only considering as the possible segmentations any segmentations remaining after removing from consideration the segmentations having evaluation results indicative of poor model fit.

4. The method of claim 1, wherein the models are selected from step functions and ramp functions.

5. The method of claim 1, wherein the step of evaluating each of the plurality of possible segmentations comprises using Bayesian Model Selection to assign weights to each segment in each segmentation, and wherein the weight associated with each segment is an accuracy measurement of the models corresponding to the segment fit to the data associated with the segment.

6. The method of claim 1, wherein the step of fitting portions of the input data corresponding to each segment in the segmentation being evaluated is performed using linear regression.

7. The method of claim 1, further comprising:

creating a treestructure having nodes representing particles corresponding to particular segmentations, where at index i, a set of particles (parent particles), each corresponding to a particular segmentation, are held active; and

for each new data item received from the input data stream at index i+1, creating a plurality of child nodes to each node active parent particle node, each child node corresponding to either a continuation of the segment to which the parent particle node corresponds or a start of a new segment with a new model.

8. The method of claim 1, further comprising

representing each possible segmentation for n data points as a plurality of particles, wherein the removal step comprises removing from consideration particles representing poor-fit segmentations.

9. The method of claim 8, further comprising upon receiving an additional data point (n+1) on the input data, spawning additional particles as child particles from each active particle and corresponding to each new possible segmentation involving the new data point.

10. The method of claim 1, wherein the step of using the output to control at least one parameter of the hydrocarbon industry process comprises:

indicating to a controller of the at least one parameter that a likely change of operating condition has occurred; and

adjusting the at least one parameter in response to the indication that a likely change of operating condition has occurred.

11. The method of claim 1, wherein generating an output from at least one of the plurality of possible segmentations and the models corresponding to the segments of the at least one of the plurality of possible segmentations comprises:

computing a probability of a first condition having occurred;

feeding the probability of the first condition having occurred into an inference engine;

operating the inference engine to determine whether a first event has occurred.

12. The method of claim 11, computing a probability of a first condition having occurred comprises computing a value as a function of the models corresponding to the segments of each segmentation under consideration, comparing that value to a threshold condition, and declaring that the probability of the first condition having occurred as the weighted sum of the probabilities associated with the segmentation for which the computed value satisfies the threshold condition.

13. The method of claim 11, further comprising:

inputting at least one additional probability of a second condition having occurred into the inference engine;

operating the inference engine to determine the probability of each of the first event and of a second event having occurred causing the first condition to occur.

14. The method of claim 11, wherein the input stream comprises pit volume data as a function of a time index, the value computed is the pit volume and first condition is whether the pit volume exceeds a given threshold, and the probability that the pit volume data exceeds a threshold and the second condition is a change in rig state, and wherein the inference engine is a Bayesian Belief Network modeling the probability of a kick occurring based on the probability of pit volume exceeding a given threshold, and the probability of a change in rig state.

15. A method for automating or partially automating a process in the hydrocarbon industry, where the hydrocarbon process is subject to a change in one or more operating conditions, can be controlled by at least one parameter and is monitored by at least one sensor providing an input data stream, comprising:

receiving the stream of input data from the at least one sensor;

segmenting the received input data stream into segments of data separated by changepoints, wherein a first segment of the segmented data stream comprises a first sequence of consecutively received data items in the input data stream where the first sequence of consecutively received data items fits into a first model from a set of predetermined models, and wherein a changepoint for the segment comprises a sequential data item commencing a second sequence of consecutively received data items for which a second model from the set of predetermined models provides a better fit with the second sequence of consecutively received data items than the first model;

generating an output from at least one of the segments of data and the changepoints; and

using the output to control at least one parameter of the hydrocarbon process.

16. The method of claim 15, wherein the second sequence of consecutively received data comprises a second segment.

17. The method of claim 15, wherein the set of predetermined models comprises a step function and a ramp function.

18. A system for automating or partially automating a process in the hydrocarbon industry, where the hydrocarbon-industry-process is subject to a change in one or more operating conditions, can be controlled by at least one parameter and is monitored by at least one sensor providing an input data stream, comprising:

a processor configured to receive the stream of input data from the at least one sensor;

software configured to segment the received input data stream into segments of data separated by changepoints, wherein a segment of the segmented data stream comprises a first sequence of consecutively received data items in the input data stream and fitting the first sequence of consecutively received data items in the segment into a first of a set of predetermined models, and wherein a changepoint for the segment comprises a sequential data item commencing a second sequence of consecutively received data items for which a second model of the set of predetermined models provide a better fit than the first predetermined model;

an output for outputting signals computed using the predetermined model corresponding to at least one of the segments of data for controlling at least one parameter of the hydrocarbon industry process.

19. A method of operating an automated drilling apparatus, comprising:

receiving a measurement indicative of depth-of-cut;

receiving weight-on-bit and drill-bit rotational speed measurements;

determining a functional relationship between rate of penetration and weight-on-bit;

from the functional relationship between depth-of-cut and weight-on-bit, determining a second functional relationship defining rate-of-penetration as a function of drill-bit rotational speed and weight-on-bit;

determining operating constraints defining a safe operating envelope as a function of drill-bit rotational speed and weight-on-bit;

determining the rotational speed and weight-on-bit parameters that provide the optimal rate-of-penetration location within the safe operating envelope; and

outputting a combination of drill-bit rotational speed and weight-on-bit to move the drill-bit rotational speed and weight-on-bit towards the rotational speed and weight-on-bit parameters for the optimal rate-of-penetration location.

20. The method of operating an automated drilling apparatus of 19, further comprising:

upon receiving additional measurements of depth-of-cut, weight-on-bit and drill-bit rotational speed, updating the functional relationship between depth-of-cut and weight-on-bit and the second functional relationship defining rate-of-penetration as a function of drill-bit rotational speed and weight-on-bit, and the operating constraints defining the safe operating envelope.

21. The method of operating an automated drilling apparatus of 20, wherein the step of updating the functional relationship defining rate-of-penetration as a function of drill-bit rotational speed and weight-on-bit comprises:

postulating that the data streams are segmented according to a plurality of possible segments divided by changepoints each indicative of a change in operating condition;

evaluating each segmentation by;

fitting the input stream data corresponding to each segment in the segmentation to a model corresponding to the each segment in the segmentation; and

evaluating the segmentations by determining how well the models for the segments of each segmentation fit the input data corresponding to each segment of each segmentation; and

using at least one of the most likely segmentations and the models corresponding to the segments of the at least one most likely segmentations to determine the functional relationship between depth-of-cut and weight-on-bit and the second functional relationship defining rate-of-penetration as a function of drill-bit rotational speed and weight-on-bit, and the operating constraints defining the safe operating envelope.

22. The method of operating an automated drilling apparatus of 21, further comprising:

removing from consideration any segmentations having low priority of providing a close-fit modeling of the data streams.

23. The method of operating an automated drilling apparatus of 21, further comprising:

upon receiving an additional data point, postulating further segmentations based on currently active segmentations and possible alternative segmentations deriving from the active segmentations wherein for each active segmentation the possible alternative segmentations represent continuation of the each active segmentation, and new segmentations representing alternative models for the received additional data point.

24. A method of operating an automated industrial process having a desired performance goal and at least one controllable parameter and at least one sensor measuring a property useful for determining a functional relationship between the desired performance goal and the at least one controllable parameter, wherein the industrial process is subject to at least one dynamic constraint, comprising:

receiving measurement data from the at least one sensor;

modeling the received measurement data to determine the functional relationship between the desired performance goal and the controllable parameter;

modeling the at least one dynamic constraint based on the received measurement data; and

using the functional relationship between the desired performance goal and the at least one dynamic constraint to determine a suggested parameter setting for the at least one controllable parameter such that according to the determined functional relationship between the at least one controllable parameter and the desired performance goal expected improved performance would be achieved using the suggested parameter setting while operating within the at least one dynamic constraint.

25. The method of operating an automated industrial process of claim 24, further comprising:

postulating that the measurement data are segmented according to a plurality of possible segments divided by changepoints each indicative of a change in operating condition;

evaluating each segmentation by;

fitting the measurement data corresponding to each segment in the segmentation to a model corresponding to the each segment in the segmentation; and

using at least one of the most likely segmentations and the models corresponding to the segments of the at least one most likely segmentations to determine the functional relationship between the desired performance goal and the at least one controllable parameter, and the at least one dynamic constraint.

26. The method for automating an industrial process of claim 24, wherein the stream of input data includes azimuth and inclination indexed by depth, the method comprising:

wherein the segmentation step comprises segmenting the azimuth and inclination data into a plurality segments each having models associated therewith for computing azimuth and inclination as a function of depth;

the method further comprising associating weights to each segmentation wherein the weights are indicative of how well the models of the segmentation fits the data stream;

wherein using the at least one of the segmentations, comprises

for all active segmentations using the models associated with the segment corresponding to a depth to compute inclination and azimuth values;

using the computed inclination and azimuth values to compute dogleg severity and toolface; and

computing a weighted average for each of the dogleg severity and toolface computations from the dogleg severity and toolface values calculated for each segmentation; and

providing the weighted average dogleg severity and toolface to a human driller or to an automated drill controller.

27. The method for automating an industrial process of claim 24, wherein each segment is defined by a first changepoint (MD1) and a second changepoint or the current depth location (MD2),

wherein dogleg severity for a particular depth location (MD2) and a particular segmentation p is computed using the formula:

DLS_p=ACOS(I2−

I1)−

SIN(I1)*SIN(I2)*(1.0−

COS(A2−

A1))/(MD2−

MD1)) 

 

(1)

y=COS(A2−

A1)*SIN(I2)*SIN(I1) 

 

(2)

GTF_p=ACOS(COS(I1)*y−

COS(I2))/(SIN(I)*SIN(ACOS(y))) 

 

(3)

where;

I1 and I2 are the inclination values computed at the first changepoint MD1 starting the segment to which the particular depth location MD2 belongs and at the particular depth location MD2 using the inclination model associated with the segment to which the particular depth location MD2 belongs, respectively

A1 and A2 are the azimuth values computed at the first changepoint MD1 starting the segment to which the particular depth location MD2 belongs and at the particular depth location MD2 using the inclination model associated with the segment to which the particular depth location MD2 belongs, respectively

DLS_p is the dogleg severity at MD2 computed with the segmentation p

GTF_p is the toolface at MD1 computed with the segmentation p; and

wherein the weighted averages for dogleg severity and toolface are computed using the weights associated with each segmentation using the formulas;

28. A hydrocarbon industry control system having at least one sensor and at least one controllable parameter, wherein the sensor produces a stream of input data indicative of operating conditions encountered by a controllable piece of equipment, comprising:

a processor with a communications connection to receive the stream of input data;

a storage system comprising processor-executable instructions that comprises instructions to cause the processor to;

upon receiving a new data item from the input data stream;

postulate that the data stream is segmented according to a plurality of possible segmentations each comprising a plurality of segments divided by changepoints each changepoint indicative of a change in operating condition;

evaluate each segmentation by;

fitting the input stream data corresponding to each segment in the segmentation to a model corresponding to the each segment in the segmentation; and

evaluating the segmentations by determining how well the models for the segments of each segmentation fit the input data corresponding to each segment of each segmentation; and

use at least one of the segmentations and the models corresponding to the segments of the at least one of the segmentations as input to a control program controlling at least one parameter of the process in the hydrocarbon industry.

29. The control system of claim 28, wherein the processor-executable instructions further comprises instructions to cause the processor to:

periodically remove from consideration segmentations having evaluation results indicative of poor model fit of the models corresponding to the segments of the segmentation; and

upon receiving additional data on the data stream only consider further segmentations based upon segmentations that remain after the removing from consideration segmentations having evaluation results indicative of poor model fit.

30. The control system of claim 28, wherein the models are selected from step functions and ramp functions.

31. The control system of claim 28, wherein the instructions to evaluate each segmentation cause the processor to perform a Bayesian Model Selection to assign weights to each segment in each segmentation wherein the weight associated with each segment is an accuracy measurement of the models corresponding to the segment fit to the data associated with the segment.

32. The control system of claim 28, wherein the fitting of the input data stream data corresponding to each segment in the segmentation is performed using linear regression.

33. The control system of claim 28, wherein the processor-executable instructions further comprises instructions to cause the processor to:

create a treestructure having nodes representing particles corresponding to particular segmentations;

to, at index i, hold active a set of particles (parent particles), each corresponding to a particular segmentation;

for each new data item received from the input data stream at index i+1, creating a plurality of child nodes to each node active parent particle node, each child node corresponding to either a continuation of the segment to which the parent particle node corresponds or a start of a new segment with a new model.

34. The control system of claim 28, wherein the processor-executable instructions further comprises instructions to cause the processor to:

representing each possible segmentation for n data points as a plurality of particles;

wherein the removal step comprises removing from consideration particles representing poor-fit segmentations; and

upon receiving an additional data point (n+1) spawning additional particles as child particles from each active particle and corresponding to each new possible segmentation involving the new data point.

35. The control system of claim 28, wherein the instruction to use at least one of the segmentations and the models corresponding to the segments of the at least one of the segmentations to a control program controlling at least one parameter of the process in the hydrocarbon industry comprise instructions to cause the processor to

indicate to a controller of the at least one parameter that a likely change of operating condition has occurred;

and

adjusting the at least one parameter in response to the indication that a likely change of operating condition has occurred.

36. The control system of claim 28, wherein the instructions to use at least one of the segmentations and the models corresponding to the segments of the at least one of the segmentations as input to a control program controlling at least one parameter of the process in the hydrocarbon industry comprises instructions to cause the processor to:

compute a probability of a first condition having occurred;

feed the probability of the first condition having occurred into an inference engine; and

operate the inference engine to determine whether a first event has occurred.

37. The control system of claim 36, wherein the instructions to compute a probability of a first condition having occurred comprises instructions to cause the processor to compute a value as a function of the models corresponding to the segments of each segmentation under consideration, comparing that value to a threshold condition, and declaring that the probability of the first condition having occurred as the weighted sum of the probabilities associated with the segmentation for which the computed value satisfies the threshold condition.

38. The control system of claim 28, wherein the processor-executable instructions further comprises instructions to cause the processor to:

input at least one additional probability of a second condition having occurred into the inference engine;

operate the inference engine to determine the probability of each of the first event and of a second event having occurred causing the first condition to occur.

39. The control system of claim 28, wherein the input stream comprises pit volume data as a function of a time index, the value computed is the pit volume and first condition is whether the pit volume exceeds a given threshold, and the probability that the pit volume data exceeds a threshold and the second condition is a change in rig state, and wherein the inference engine is a Bayesian Belief Network modeling the probability of a kick occurring based on the probability of pit volume exceeding a given threshold, and the probability of a change in rig state.