Method and system for updating a model in a model predictive controller

Patent metadata

Field	Value
Publication number	US-9581980-B2
Application number	US-201213654698-A
Country	US
Kind code	B2
Filing date	Oct 18, 2012
Priority date	Apr 19, 2010
Publication date	Feb 28, 2017
Grant date	Feb 28, 2017

Abstract

Exemplary embodiments relate to a method and system for updating a model in a model predictive controller. The system executing a method that includes assessing the deviation of the operating performance level from the desired performance level of the process plant, and diagnosing the model predictive control for the model plant mismatch by updating the model in a model predictive controller. The step of diagnosing the model predictive controller includes determining the model prediction error in relation to model plant mismatch, quantifying the model plant mismatch, and updating the model in the model predictive controller.

First claim

What is claimed is:

1. A method for updating a model in a model predictive controller of a distributed control system in a process plant, the method comprising:
assessing deviation of operating performance level from desired performance level of the process plant;
diagnosing model predictive control (MPC) for model plant mismatch (MPM), for updating the model in the model predictive controller and of the MPC thereof; wherein diagnosing said MPC comprises:
a. determining a model prediction error in relation to MPM, wherein determining the model prediction error in relation to MPM comprises:
adding white noise to set point values of controllers in the process plant;
providing set points with white noise to the controllers in the process plant and to the model of the MPC; and
estimating the model prediction error by calculating a difference between output values of the process plant and of the model of the MPC; and

b. quantifying the MPM using non-zero lag correlation coefficients between the model prediction error and each of the manipulated variables of said MPC, said manipulated variables of the MPC being the set point values with added white noise of the controllers in the process plant; and
c. updating the model in said model predictive controller for operation of the process plant based on the quantified MPM.

2. The method as claimed in claim 1 , wherein assessing the deviation of the operating performance level from the desired performance level of the process plant includes monitoring the performance of the MPC, and detecting the cause for said deviation, said deviation of the operating performance level from the desired performance level of the process plant corresponds to MPM.

3. The method as claimed in claim 1 , wherein diagnosing said MPC includes correcting said deviation of the operating performance level from the desired performance level of the process plant by correcting the model of said MPC for MPM and of the MPC thereof.

4. The method as claimed in claim 1 , wherein the model update is performed offline or online.

5. The method as claimed in claim 1 , wherein the process involved in the process plant is open loop or closed loop.

6. The method as claimed in claim 1 , wherein said model includes an entire model of the process plant and/or a sub model thereof.

7. A system for updating a model in a model predictive controller (MPC) according to the method of claim 1 , the system comprising:
a performance monitor unit for monitoring performance of the MPC;
a detection unit for detecting a cause for deviation of an operating performance level from a desired performance level of the process plant;
a white noise generator unit for generating white noise signals sufficient to compute non-zero lag correlation coefficients;
an estimator unit for estimating model prediction error by calculating a difference between an output value of the process plant and of an existing model of the MPC;
a quantifier unit for quantifying a model plant mismatch (MPM) from the estimated model prediction error; and
an updater unit for providing an updated model based on the quantified MPM and updating the model of the process plant in the model predictive controller.

8. A system for updating a model in a model predictive controller (MPC) of a distributed control system in a process plant, the system comprising:
a performance monitor unit that monitors performance of the MPC;
a detection unit that detects the cause for deviation of an operating performance level from a desired performance level of the process plant;
a white noise generator unit that generates white noise signals sufficient to compute non-zero lag correlation coefficients;
an estimator unit that adds white noise to set point values of the MPC of the process plant and estimates a model prediction error by calculating a difference between an output value of the process plant and of an existing model of the MPC;
a quantifier unit that quantifies a model plant mismatch (MPM) from the estimated model prediction error and provides set points with white noise to controllers in the process plant and to the model of the MPC; and
an updater unit that provides an updated model based on the quantified MPM and updates the model of the process plant in the model predictive controller.

9. A non-transitory computer readable medium having computer program code stored thereon, which when the non-transitory computer readable medium is in communicable contact with a computer, the computer program code causes the computer to execute a method of updating a model in a model predictive controller, the method comprising:
assessing a deviation of an operating performance level from a desired performance level of a process plant;
diagnosing model predictive control (MPC) for a model plant mismatch (MPM) by updating the model in the model predictive controller and of the MPC thereof; wherein diagnosing said MPC comprises;
a. determining a model prediction error in relation to MPM, wherein determining the model prediction error in relation to MPM comprises:
adding white noise to set point values of controllers in the process plant;
providing set points with white noise to the controllers in the process plant and to the model of the MPC; and
estimating the model prediction error by calculating a difference between output values of the process plant and of the model of the MPC; and

b. quantifying the MPM using non-zero lag correlation coefficients between said model prediction errors and each of the manipulated variables of the MPC, said manipulated variables of the MPC being the set point values with added white noise of the controllers in the process plant; and
c. updating the model in said model predictive controller based on the quantified MPM.

Assignees

• Abb Research Ltd
• Abb Schweiz Ag

Inventors

• Badwe Abhijit
• Kubal Nandkishor
• Bhat Shrikant
• Nandola Naresh Naranbhai

Classifications

• G05B13/048 Primary

  using a predictor · CPC title

• G05B13/042

  in which a parameter or coefficient is automatically adjusted to optimise the performance · CPC title

• G05B13/041

  in which a variable is automatically adjusted to optimise the performance · CPC title