Model predictive controller solution analysis process

• US 7,949,417 B2
• Filed: 09/22/2006
• Issued: 05/24/2011
• Est. Priority Date: 09/22/2006
• Status: Active Grant

• Current Assignee: ExxonMobil Research & Engineering Company (Exxon Mobil Corporation)
• Original Assignee: ExxonMobil Research & Engineering Company (Exxon Mobil Corporation)
• Inventors: Wang, Robert K., Barrett-Payton, Dave, Punuru, Adi R., Peterson, Tod J., Emigholz, Kenneth F.
• Primary Examiner(s): Decady; Albert
• Assistant Examiner(s): Stevens; Thomas H
• Application Number:
  US11/525,221
• Publication Number:
  US 20080077257A1
• Time in Patent Office: 1,705 Days
• Field of Search: 700/44, 700/53, 700/28, 700/36, 700/33, 700/34
• US Class Current: 700/44
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

The solution from a multivariable predictive controller (MPC) is analyzed and described by providing quantitative input to operators regarding the effect of changing controller limits on the MPC controller solution. This information allows a rapid operator response to changes and more optimal process operation.

Claims

1. A method of analyzing a solution from a multivariable predictive controller, comprising:

obtaining a solution from a multivariable predictive controller having a steady-state optimizer that results in different variable constraint statuses, wherein the solution includes controlled variables that are predicted from manipulated variables; and

operating on the solution to obtain a relationship between constrained variables and unconstrained variables to determine how unconstrained variables respond to changes in constrained variables; and

for each constrained variable, determining how far it can be moved until a next constraint is reached; and

using the relationship between the constrained and unconstrained variables to determine an amount of change needed to move a constraint for a violated variable such that the violated variable is no longer constrained into feasibility.

2. The method of claim 1, wherein the solution is represented in a matrix and operating on the solution includes pivoting the constrained controlled variables with the unconstrained manipulated variables in the matrix.

3. The method of claim 1, wherein the solution is represented in a matrix formed by simulating the multivariable predictive controller and perturbing each constraint a predetermined amount, wherein a ratio of a change of unconstrained variable to the predetermined amount is defined as a gain for each constrained and unconstrained variable pair.

4. The method of claim 1, further comprising using the relationship between the constrained and unconstrained variables to calculate an amount of change needed in a constrained variable to reach a next constraint in an unconstrained or a constrained variable.

5. The method of claim 4, further comprising determining a set of limits for each variable and selecting a limit to use for calculating the amount of change.

6. The method of claim 1, further comprising using the relationship between the constrained and unconstrained variables to determine how much change in a constraint affects an unconstrained variable.

7. A method of operating a control system for use with a process facility, comprising:

extracting a raw gain matrix from a base model file, including manipulated variables and controlled variables related to the process, based on a steady-state response between the manipulated variables and the controlled variables;

classifying the manipulated variables and controlled variables by an active constraint condition, wherein the classification is based on constrained, unconstrained or violated conditions;

calculating an amount of possible movement for each variable based on the active constraint condition classification;

wherein calculating the amount of possible movement for a constrained variable includes determining which constraint condition will be reached next;

wherein calculating the amount of possible movement for an unconstrained variable results in an operator high limit, an operator low limit or a step limit;

changing the order of the gain matrix based on the active constraint condition classification to obtain a model matrix representative of an optimization solution;

forming a result matrix by pivoting the constrained controlled variables with the unconstrained manipulated variables in the model matrix to form the result matrix; and

,

using the result matrix and the amount of possible movement to calculate the response of the unconstrained variables to changes in the constrained variables.

8. The method of claim 7, wherein each variable has a set of limits and calculating the amount of possible movement includes selecting a limit from the set of limits to use in the calculation.

9. The method of claim 7, wherein changing the order of the gain matrix includes moving unconstrained manipulated variables to the top of gain matrix and constrained controlled variables to the left of gain matrix.

10. The method of claim 7, wherein changing the order of the gain matrix results in a matrix having (i) a model of unconstrained manipulated variables to constrained controlled variables, (ii) a model of constrained manipulated variables to constrained controlled variables, (iii) a model of unconstrained manipulated variables to unconstrained controlled variables, and (iv) a model of constrained manipulated variables to unconstrained controlled variables.

11. The method of claim 7, wherein using the result matrix includes determining how far a constrained variable can be changed before a next constraint is reached.

12. The method of claim 7, wherein using the result matrix includes determining all constrained variables that affect each unconstrained variable.

13. The method of claim 7, wherein using the result matrix includes determining how much a change in each constraint affects an associated unconstrained variable.

14. The method of claim 7, wherein using the result matrix includes calculating movement required to achieve feasibility for each constraint.

15. A control system for use with a process, comprising:

a storage device that stores a base model file including manipulated variables and controlled variables related to the process; and

a controller associated with the storage device that extracts a raw gain matrix from the base model file based on a steady-state response between the manipulated variables and the controlled variables, wherein the controller uses an optimization solution to describe how unconstrained variables respond to changes in constrained variables by

classifying the manipulated variables and controlled variables by an active constraint condition, wherein the classification is based on constrained, unconstrained or violated conditions,

calculating an amount of possible movement for each variable based on the active constraint condition classification,

wherein the controller calculates the amount of possible movement for a violated variable by determining what movement will return the variable to a limit;

changing the order of the raw gain matrix based on the active constraint condition to obtain a model matrix representative of an optimization solution,

forming a result matrix by pivoting the constrained controlled variables with the unconstrained manipulated variables in the model matrix to form the result matrix, and

using the result matrix and the amount of possible movement to calculate the response of the unconstrained variables to changes in the constrained variables.

16. The control system of claim 15, wherein the controller calculates the amount of possible movement for an unconstrained variable by evaluating possible change in a result of an optimization function.

17. The control system of claim 16, wherein the result of the optimization function is an operator high limit, an operator low limit or a step limit.

18. The control system of claim 15, wherein the controller changes the order of the gain matrix by moving unconstrained manipulated variables to the top of the matrix and constrained controlled variables to the left of the matrix.

19. The control system of claim 15, wherein the controller changes the order of the gain matrix to result in a matrix having (i) a model of unconstrained manipulated variables to constrained controlled variables, (ii) a model of constrained manipulated variables to constrained controlled variables, (iii) a model of unconstrained manipulated variables to unconstrained controlled variables, and (iv) a model of constrained manipulated variables to unconstrained controlled variables.

20. The control system of claim 15, wherein the controller determines how far a constrained variable can be changed before another constraint is reached by using the result matrix.

21. The control system of claim 15, wherein the controller determines all constrained variables that affect each unconstrained variable by using the result matrix.

22. The control system of claim 15, wherein the controller determines how much a change in each constraint affects an associated unconstrained variable by using the result matrix.

23. The control system of claim 15, wherein the controller calculates movement required to achieve a non-violated variable for each constraint by using the result matrix.

24. A method of operating a control system for use with a process facility, comprising:

extracting a raw gain matrix from a base model file, including manipulated variables and controlled variables related to the process, based on a steady-state response between the manipulated variables and the controlled variables;

classifying the manipulated variables and controlled variables by an active constraint condition, wherein the classification is based on constrained, unconstrained or violated conditions;

calculating an amount of possible movement for each variable based on the active constraint condition classification;

wherein calculating the amount of possible movement for a constrained variable includes determining which constraint condition will be reached next;

wherein calculating the amount of possible movement for a violated variable includes determining what movement will return the variable to a limit;

changing the order of the raw gain matrix based on the active constraint condition to obtain a model matrix representative of an optimization solution;

forming a result matrix by pivoting the constrained controlled variables with the unconstrained manipulated variables in the model matrix to form the result matrix; and

,

using the result matrix and the amount of possible movement to calculate the response of the unconstrained variables to changes in the constrained variables.

25. The method of claim 24, wherein calculating the amount of possible movement for an unconstrained variable results in an operator high limit, and operator low limit or a step limit.

26. The method of claim 24, wherein each variable has a set of limits and calculating the amount of possible movement includes selecting a limit from the set of limits to use in the calculation.

27. The method of claim 24, wherein changing the order of the gain matrix includes moving unconstrained manipulated variables to the top of the gain matrix and constrained controlled variables to the left of the gain matrix.

28. The method of claim 24, wherein changing the order of the gain matrix results in a matrix having (i) a model of unconstrained manipulated variables to constrained controlled variables, (ii) a model of constrained manipulated variables to constrained controlled variables, (iii) a model of unconstrained manipulated variables to unconstrained controlled variables, and (iv) a model of constrained manipulated variables to unconstrained controlled variables.

29. The method of claim 24, wherein using the result matrix includes determining how far a constrained variable can be changed before a next constraint is reached.

30. The method of claim 24, wherein using the result matrix includes determining all constrained variables that affect each unconstrained variable.

31. The method of claim 24, wherein using the result matrix includes determining how much a change in each constraint affects an associated unconstrained variable.