On-Line Adaptive Model Predictive Control in a Process Control System

• US 20090143872A1
• Filed: 11/07/2008
• Published: 06/04/2009
• Est. Priority Date: 09/30/2005
• Status: Active Grant

• Current Assignee: Fisher-Rosemount Systems Incorporated (Emerson Electric Company)
• Original Assignee: Fisher-Rosemount Systems Incorporated (Emerson Electric Company)
• Inventors: Wojsznis, Wilhelm K., Thiele, Dirk
• Granted Patent: US 7,856,281 B2
• Time in Patent Office: Days
• US Class Current: 700/30
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

A method of creating and using an adaptive DMC type or other MPC controller includes using a model switching technique to periodically determine a process model, such as a parameterized process model, for a process loop on-line during operation of the process. The method then uses the process model to generate an MPC control model and creates and downloads an MPC controller algorithm to an MPC controller based on the new control model while the MPC controller is operating on-line. This technique, which is generally applicable to single-loop MPC controllers and is particularly useful in MPC controllers with a control horizon of one or two, enables an MPC controller to be adapted during the normal operation of the process, so as to change the process model on which the MPC controller is based to thereby account for process changes. The adaptive MPC controller is not computationally expensive and can therefore be easily implemented within a distributed controller of a process control system, while providing the same or in some cases better control than a PID controller, especially in dead time dominant process loops, and in process loops that are subject to process model mismatch within the process time to steady state.

Claims

1. An adaptive process controller for controlling one or more devices within a process plant, the adaptive process controller comprising:

a controller having a prediction unit that uses a prediction model to determine a prediction signal for a set of one or more process variables, a prediction error unit that combines the prediction signal with a set point signal to produce a prediction error signal, and a combiner that multiplies the prediction error signal by a controller gain signal to produce a change in a control signal and that uses the change in the control signal to develop a control signal for controlling one of the set of process variables;

a process model estimation unit communicatively coupled to one or more devices within the process plant to collect process data during the on-line operation of the process plant and that determines a new process model representing the operation of a portion of the process plant from the collected process data, the new process model including a set of model parameters defining characteristics of the portion of the process plant; and

a controller adaptation unit that uses a closed form equation that expresses a new controller gain signal as a function of at least one of the set of process model parameters of the new process model to calculate the new controller gain signal and that adapts the controller to use the new controller gain signal while the controller is operating on-line to control the portion of the process plant.

2. The adaptive controller of claim 1, wherein the process model estimation unit determines a process dead time and a process time constant as elements of the set of model parameters.

3. The adaptive controller of claim 1, wherein the controller adaptation unit further determines a prediction horizon for the controller from the set of model parameters.

4. The adaptive controller of claim 1, wherein the controller adaptation unit determines a controller execution rate from the set of model parameters and provides the controller execution rate to the controller to set the execution rate of the controller during subsequent operation of the controller in controlling the portion of the process plant.

5. The adaptive controller of claim 1, wherein the controller adaptation unit determines a step response model from one or more of the model parameters, a controller execution time and a prediction horizon.

6. The adaptive controller of claim 1, wherein the controller adaptation unit determines a step response model from one or more of the model parameters, a controller execution time and a prediction horizon, the step response model including a gain vector defining a set of gains for each of a set of response times, and wherein the closed form equation expresses the new controller gain signal as a function of the set of gains of the step response model and a penalty on move.

7. The adaptive controller of claim 6, wherein the closed form equation expresses the new controller gain signal based on a control horizon equal to two controller scan periods.

8. The adaptive controller of claim 6, wherein the closed form equation expresses the new controller gain signal based on a control horizon equal to one controller scan period.

9. The adaptive controller of claim 6, wherein the closed form equation expresses the new controller gain signal based on a control horizon that is at least ten times shorter than a prediction horizon used by the controller.

10. The adaptive controller of claim 6, wherein the controller adaptation unit calculates the penalty on move as a function of one or more of the process model parameters.

11. The adaptive controller of claim 1, wherein the process model parameters include a process dead time parameter and a process gain parameter and wherein the controller adaptation unit calculates the penalty on move as a function of the process dead time parameter and the process gain parameter.

12. The adaptive controller of claim 1, wherein the controller adaptation unit develops a new prediction model from the new process model and provides the new prediction model to the controller for use in the prediction unit of the controller as the prediction model during subsequent operation of the controller.

13. The adaptive controller of claim 1, wherein the controller adaptation unit further determines a time to steady state from the process model and determines a combination of a prediction horizon and an execution rate for the controller based on the time to steady state.

14. The adaptive controller of claim 1, wherein the new controller gain signal comprises a gain vector.

15. The adaptive controller of claim 1, wherein the new controller gain signal comprises a scalar value.

16. A method of adapting a process controller that controls one or more devices to implement a process within a process plant using a prediction unit that uses a prediction model to determine a prediction signal for a set of one or more process variables, a prediction error unit that combines the prediction signal with a set point signal to produce a prediction error signal, and a combiner that multiplies the prediction error signal with a controller gain signal to produce a change in a control signal, and that uses the change in the control signal to develop a control signal for controlling the one or more devices, the method comprising:

collecting process data during on-line operation of the process plant;

determining, from the collected process data, a new process model including a set of model parameters that characterize the process;

using a closed form equation that expresses a new controller gain signal as a function of at least one of the set of model parameters to calculate the new controller gain signal; and

providing the new controller gain signal to the process controller for use as the controller gain signal in controlling a portion of the process plant during subsequent operation of the process controller.

17. The method of claim 16, wherein determining the set of model parameters includes determining a process dead time and a process time constant from the collected process data.

18. The method of claim 16, further including determining a prediction horizon from the set of model parameters.

19. The method of claim 16, further including determining a controller execution rate from the set of model parameters and wherein providing the new controller gain signal to the process controller includes providing the controller execution rate to the process controller to set the execution rate of the process controller during subsequent operation of the process controller.

20. The method of claim 16, further including determining a step response model from one or more of the set of model parameters, a controller execution rate and a prediction horizon.

21. The method of claim 16, further including determining a step response model from one or more of the set of model parameters, a controller execution rate and a prediction horizon, the step response model including a gain vector defining a set of gains for each of a set of response times, and wherein using the closed form equation to calculate a new controller gain signal includes using a closed form equation that is a function of the set of gains of the step response model and a penalty on move.

22. The method of claim 21, wherein the closed form equation has a form based on a control horizon equal to two controller scan periods.

23. The method of claim 21, wherein the closed form equation has a form based on a control horizon equal to one controller scan period.

24. The method of claim 21, wherein the closed form equation has a form based on a control horizon that is at least ten times shorter than a prediction horizon used by the process controller.

25. The method of claim 21, further including calculating the penalty on move as a function of one or more of the set of model parameters.

26. The method of claim 21, wherein the set of model parameters includes a process dead time parameter and a process gain parameter and wherein calculating the penalty on move includes calculating the penalty on move as a function of the process dead time parameter and the process gain parameter.

27. The method of claim 16, further including developing a new prediction model from the new process model and providing the new prediction model to the process controller for use in the prediction unit of the process controller as the prediction model during subsequent operation of the process controller.

28. The method of claim 16, further including determining a time to steady state from the process model and determining a combination of a prediction horizon and an execution rate for the controller based on the time to steady state.

29. The method of claim 16, wherein the new controller gain signal comprises a gain vector.

30. The method of claim 16, wherein the new controller gain signal comprises a scalar value.

31. The method of claim 16, wherein the new controller gain signal comprises a multi-dimensional matrix.