System and methodology and adaptive, linear model predictive control based on rigorous, nonlinear process model

• US 6,826,521 B1
• Filed: 04/06/2000
• Issued: 11/30/2004
• Est. Priority Date: 04/06/2000
• Status: Expired due to Term

• Current Assignee: ABB Automation GmbH (ABB Limited)
• Original Assignee: ABB Automation GmbH (ABB Limited)
• Inventors: Kroll, Andreas, Hess, Todd M., Modén, Per Erik, Lindberg, Carl-Fredrik S. M., Praprost, Kenneth L., Petersson, Mikael
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): Thangavelu, Kandasamy
• Application Number:
  US09/544,390
• Time in Patent Office: 1,699 Days
• Field of Search: 703/2, 703/12, 703/13, 700/29, 700/38, 700/44, 700/45, 702/56, 706/21
• US Class Current: 703/12
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

A methodology for process modeling and control and the software system implementation of this methodology, which includes a rigorous, nonlinear process simulation model, the generation of appropriate linear models derived from the rigorous model, and an adaptive, linear model predictive controller (MPC) that utilizes the derived linear models. A state space, multivariable, model predictive controller (MPC) is the preferred choice for the MPC since the nonlinear simulation model is analytically translated into a set of linear state equations and thus simplifies the translation of the linearized simulation equations to the modeling format required by the controller. Various other MPC modeling forms such as transfer functions, impulse response coefficients, and step response coefficients may also be used. The methodology is very general in that any model predictive controller using one of the above modeling forms can be used as the controller. The methodology also includes various modules that improve reliability and performance. For example, there is a data pretreatment module used to pre-process the plant measurements for gross error detection. A data reconciliation and parameter estimation module is then used to correct for instrumentation errors and to adjust model parameters based on current operating conditions. The full-order state space model can be reduced by the order reduction module to obtain fewer states for the controller model. Automated MPC tuning is also provided to improve control performance.

Claims

1. A method for controlling a process comprising:

a) receiving plant measurement variables from a regulatory control system;

b) applying said plant measurement variables to update one or more variables of a nonlinear model;

c) linearizing said updated nonlinear model when a change in said one or more of said model variables has exceeded an associated predetermined threshold; and

d) passing a MPC format model converted from said linearized model to a model predictive controller.

2. A method for controlling a process comprising:

a) receiving plant measurement variables from a regulatory control system;

b) applying said plant measurement variables to update one or more variables of a nonlinear model;

c) linearizing said updated nonlinear model; and

d) passing a MPC format model converted from said linearized model to a model predictive controller,

said updated nonlinear model linearized when one or more model prediction errors in said MPC format model currently operational in said model predictive controller has exceeded an associated predetermined threshold.

3. A method for controlling a process comprising:

a) applying simulation stimuli to update one or more variables of a nonlinear model comprising;

(i) pretreating said simulation stimuli;

(ii) reconciling said pretreated simulation stimuli; and

(iii) using said reconciled and pretreated simulation stimuli to update said nonlinear model;

b) linearizing said updated nonlinear model when a change in said one or more of said model variables has exceeded an associated predetermined threshold;

c) converting said linearized model to a full order state space model;

d) creating from said full order state space model a state space model having fewer states than said full order state space model;

e) converting said fewer states state space model to a MPC format model; and

f) evaluating the performance of said MPC format model with the tuning for a presently existing model of said process in a model predictive controller versus the performance of said presently existing model with said tuning and either;

passing said MPC format model with said presently existing model tuning to a model predictive controller when said performance evaluation of said MPC format model exceeds a first predetermined limit;

or

computing new tuning for said MPC format model when said performance evaluation falls below said first predetermined limit and repeating said evaluations;

or

returning said MPC format model to said creating a MPC format model having fewer states than said full order state space model to change the number of states in said MPC format model when said performance of said MPC format model falls below said first predetermined limit.

4. A method for controlling a process comprising:

a) applying simulation stimuli to update one or more variables of a nonlinear model comprising;

(i) pretreating said simulation stimuli;

(ii) reconciling said pretreated simulation stimuli; and

(iii) using said reconciled and pretreated simulation stimuli to update said nonlinear model;

b) linearizing said updated nonlinear model when a change in said one or more model prediction errors in a MPC format model currently operational in a model predictive controller has exceeded an associated predetermined threshold;

c) converting said linearized model to a full order state space model;

d) creating from said full order state space model a state space model having fewer states than said full order state space model;

e) converting said fewer states state space model to a MPC format model; and

f) evaluating the performance of said MPC format model with the tuning for a presently existing model of said process in a model predictive controller versus the performance of said presently existing model with said tuning and either;

passing said MPC format model with said presently existing model tuning to a model predictive controller when said performance evaluation of said MPC format model exceeds a first predetermined limit;

or

computing new tuning for said MPC format model when said performance evaluation falls below said first predetermined limit and repeating said evaluations;

or

returning said MPC format model to said creating a MPC format model having fewer states than said full order state space model to change the number of states in said MPC format model when said performance of said MPC format model falls below said first predetermined limit.

5. A method for controlling a process comprising:

a) applying simulation stimuli to update one or more variables of a nonlinear model comprising;

(i) pretreating said simulation stimuli;

(ii) reconciling said pretreated simulation stimuli; and

(iii) using said reconciled and pretreated simulation stimuli to update said nonlinear model;

b) linearizing said updated nonlinear model when a change in said one or more of said model variables has exceeded an associated predetermined threshold;

c) converting said linearized model to a MPC format model; and

d) passing said MPC format model to a model predictive controller comprising;

evaluating the performance of said MPC format model with the tuning for a presently existing model of said process in a model predictive controller versus the performance of said presently existing model with said tuning and either;

passing said MPC format model with said presently existing model tuning to a model predictive controller when said performance evaluation of said MPC format model exceeds a first predetermined limit;

or

computing new tuning for said MPC format model when said performance evaluation falls below said first predetermined limit and repeating said evaluations.

6. A method for controlling a process comprising:

a) applying simulation stimuli to update one or more variables of a nonlinear model comprising;

(i) pretreating said simulation stimuli;

(ii) reconciling said pretreated simulation stimuli; and

(iii) using said reconciled and pretreated simulation stimuli to update said nonlinear model;

b) linearizing said updated nonlinear model when a change in said one or more model prediction errors in a MPC format model currently operational in a model predictive controller has exceeded an associated predetermined threshold;

c) converting said linearized model to a MPC format model; and

d) passing said MPC format model converted from said linearized model to a model predictive controller comprising;

evaluating the performance of said MPC format model with the tuning for a presently existing model of said process in said model predictive controller versus the performance of said presently existing model with said tuning and either;

passing said MPC format model with said presently existing model tuning to a model predictive controller when said performance evaluation of said MPC format model exceeds a first predetermined limit;

or

computing new tuning for said MPC format submodel when said performance evaluation falls below said first predetermined limit and repeating said evaluations.

7. A method for controlling a process comprising:

a) receiving plant measurement variables from a regulatory control system;

b) pretreating said plant measurement variables;

c) reconciling said pretreated plant measurement variables;

d) using said reconciled and pretreated plant measurement variables to update one or more variables of each submodel of a nonlinear model, said nonlinear model having two or more of said submodels, each of said two or more submodels having a predetermined one of two or more model predictive controllers associated therewith;

e) converting at least one updated submodel of said updated nonlinear model to a linear submodel when a change in said one or more of said updated submodel variables has exceeded a predetermined threshold, said linear submodel for operating said associated one of said two or more controllers;

f) using said nonlinear model in a real time optimizer to compute targets for all of said two or more model predictive controllers, a predetermined subset of said computed targets associated with a respective one of said two or more controllers;

g) passing each of said predetermined subsets of said computed targets associated with a respective one of said two or more model predictive controllers to said associated one of said two or more controllers;

h) converting said at least one linearized submodel to a full order state space submodel;

i) creating from said full order state space submodel a state space submodel having fewer states than said full order state space submodel;

j) converting said fewer states state space submodel to a MPC format submodel; and

k) evaluating the performance of said MPC format submodel with the tuning for a presently existing submodel of said process in said associated one of said two or more model predictive controllers versus the performance of said presently existing submodel with said tuning and either;

passing said MPC format submodel with said presently existing submodel tuning to said associated one of said two or more model predictive controllers when said performance evaluation of said MPC format submodel exceeds a first predetermined limit;

or

computing new tuning for said MPC format submodel when said performance evaluation of said MPC format submodel falls below said first predetermined limit and repeating said evaluations;

or

returning said MPC format submodel to said creating a MPC format submodel having fewer states than said full order state space submodel to change the number of states in said MPC format submodel when said performance of said MPC format submodel falls below said first predetermined limit.

8. A method for controlling a process comprising:

a) receiving plant measurement variables from a regulatory control system;

b) pretreating said plant measurement variables;

c) reconciling said pretreated plant measurement variables;

d) using said reconciled and pretreated plant measurement variables to update one or more variables of each submodel of a nonlinear model, said nonlinear model having two or more of said submodels, each of said two or more submodels having a predetermined one of two or more model predictive controllers associated therewith;

e) converting at least one updated submodel of said updated nonlinear model to a linear submodel when a change in one or more model prediction errors in an associated one of one or more MPC format submodels currently operational in an associated one of said two or more model predictive controllers has exceeded a predetermined threshold, said linear submodel for operating said associated one of said two or more controllers;

f) using said nonlinear model in a real time optimizer to compute targets for all of said two or more model predictive controllers, a predetermined subset of said computed targets associated with a respective one of said two or more controllers;

g) passing each of said predetermined subsets of said computed targets associated with a respective one of said two or more model predictive controllers to said associated one of said two or more controllers; and

h) converting said at least one linearized submodel to a full order state space submodel;

i) creating from said full order state space submodel a state space submodel having fewer states than said full order state space submodel;

j) converting said fewer states state space submodel to said MPC format submodel; and

k) evaluating the performance of said MPC format submodel with the tuning for a presently existing submodel of said process in said associated one of said two or more model predictive controllers versus the performance of said presently existing submodel with said tuning and either;

passing said MPC format submodel with said presently existing submodel tuning to said associated one of said two or more model predictive controllers when said performance evaluation of said MPC format submodel exceeds a first predetermined limit;

or

computing new tuning for said MPC format submodel when said performance evaluation of said MPC format submodel falls below said first predetermined limit and repeating said evaluations;

or

returning said MPC format submodel to said creating a MPC format submodel having fewer states than said full order state space submodel to change the number of states in said MPC format submodel when said performance of said MPC format submodel falls below said first predetermined limit.

9. A method for controlling a process comprising:

a) receiving plant measurement variables from a regulatory control system;

b) pretreating said plant measurement variables;

c) reconciling said pretreated plant measurement variables;

d) using said reconciled and pretreated plant measurement variables to update one or more variables of each submodel of a nonlinear model, said nonlinear model having two or more of said submodels, each of said two or more submodels having a predetermined one of two or more model predictive controllers associated therewith;

e) converting at least one updated submodel of said updated nonlinear model to a linear submodel when a change in said one or more of said updated submodel variables has exceeded a predetermined threshold, said linear submodel for operating said associated one of said two or more controllers;

f) using said nonlinear model in a real time optimizer to compute targets for all of said two or more model predictive controllers, a predetermined subset of said computed targets associated with a respective one of said two or more controllers;

g) passing each of said predetermined subsets of said computed targets associated with a respective one of said two or more model predictive controllers to said associated one of said two or more controllers; and

h) passing said linear model to said associated one of said two or more controllers comprising;

evaluating the performance of said MPC format submodel with the tuning for a presently existing submodel of said process in said associated one of said two or more model predictive controllers versus the performance of said presently existing submodel with said tuning and either;

passing said MPC format submodel with said presently existing submodel tuning to said associated one of said two or more model predictive controllers when said performance evaluation of said MPC format submodel exceeds a first predetermined limit;

or

computing new tuning for said MPC format submodel when said performance evaluation of said MPC format submodel falls below said first predetermined limit and repeating said evaluating.

10. A method for controlling a process comprising:

a) receiving plant measurement variables from a regulatory control system;

b) pretreating said plant measurement variables;

c) reconciling said pretreated plant measurement variables;

d) using said reconciled and pretreated plant measurement variables to update one or more variables of each submodel of a nonlinear model, said nonlinear model having two or more of said submodels, each of said two or more submodels having a predetermined one of two or more model predictive controllers associated therewith;

e) converting at least one updated submodel of said updated nonlinear model to a linear submodel when a change in one or more model prediction errors in an associated one of one or more MPC format submodels currently operational in an associated one of said two or more model predictive controllers has exceeded a predetermined threshold, said linear submodel for operating said associated one of said two or more controllers;

f) using said nonlinear model in a real time optimizer to compute targets for all of said two or more model predictive controllers, a predetermined subset of said computed targets associated with a respective one of said two or more controllers;

g) passing each of said predetermined subsets of said computed targets associated with a respective one of said two or more model predictive controllers to said associated one of said two or more controllers; and

h) passing said linear model to said associated one of said two or more controllers comprising;

evaluating the performance of said MPC format submodel with the tuning for a presently existing submodel of said process in said associated one of said two or more model predictive controllers versus the performance of said presently existing submodel with said tuning and either;

passing said MPC format submodel with said presently existing submodel tuning to said associated one of said two or more model predictive controllers when said performance evaluation of said MPC format submodel exceeds a first predetermined limit;

or

computing new tuning for said MPC format submodel when said performance evaluation of said MPC format submodel falls below said first predetermined limit and repeating said evaluating.