Extended horizon adaptive block predictive controller with an efficient prediction system
First Claim
1. A control method including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive :
- integers, the control method comprising;
receiving variables including at least one of a set of dynamic system parameters and a set of system outputs y(t) from said multivariable dynamic system;
generating a block of k filtered output z(t) predictions of said multivariable dynamic system by;
selectively filtering the set of system outputs v(t);
constructing a block state structure from the variables of said multivariable dynamic system;
using the block state structure, generating at least one of a block of k output y(t) predictions when the filtering of said set of system outputs y(t) is not performed and a block of k filtered output ψ
(t) predictions when the selective filtering of said set of system outputs y(t) is performed, each of dimension N consisting of output y(t) predictions and filtered output ψ
(t) predictions respectively at each future time step up to a prediction horizon k; and
,using i) the-block of k output predictions when the selective filtering of said set of system outputs y(t) is not performed and ii) the block of k filtered output ψ
(t) predictions when the selective filtering of said set of system outputs y(t) is performed, to generate said block of k filtered output z(t) predictions;
combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and
,outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system.
1 Assignment
0 Petitions
Accused Products
Abstract
An extended horizon adaptive block predictive controller includes a block control system and an adaptive block prediction system. The subject adaptive block prediction system includes means and procedures for accomplishing a block of k output predictions, consisting of output predictions at each future time step up to the prediction horizon k, for dynamic systems with N outputs, M inputs and model parameter order n. The prediction system is organized for performing the calculation of a k-step ahead predictor in a highly computationally and memory efficient manner. The subject prediction system, in particular, avoids computation of a large number of the predictor parameters associated with the prior art Diophantine equation method, so that memory requirements and computational complexity are significantly reduced. The subject block control system in conjunction with an adaptive block prediction system includes means for accomplishing predictive controls of dynamic systems. It subsumes most, if not all, existing extended horizon predictive controllers. The subject block control system is based on the use of a special computational method. A hardware implementation of the subject extended horizon predictive control system includes a computing means, memory means, and dynamic system input and output signal processing means. The subject invention is primarily devised for real time control of dynamic systems.
-
Citations
34 Claims
-
1. A control method including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive :
- integers, the control method comprising;
receiving variables including at least one of a set of dynamic system parameters and a set of system outputs y(t) from said multivariable dynamic system; generating a block of k filtered output z(t) predictions of said multivariable dynamic system by; selectively filtering the set of system outputs v(t); constructing a block state structure from the variables of said multivariable dynamic system; using the block state structure, generating at least one of a block of k output y(t) predictions when the filtering of said set of system outputs y(t) is not performed and a block of k filtered output ψ
(t) predictions when the selective filtering of said set of system outputs y(t) is performed, each of dimension N consisting of output y(t) predictions and filtered output ψ
(t) predictions respectively at each future time step up to a prediction horizon k; and
,using i) the-block of k output predictions when the selective filtering of said set of system outputs y(t) is not performed and ii) the block of k filtered output ψ
(t) predictions when the selective filtering of said set of system outputs y(t) is performed, to generate said block of k filtered output z(t) predictions;combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and
,outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20)
- integers, the control method comprising;
-
21. A system controller, including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive integers comprising:
-
means adapted for receiving variables from an associated multivariable dynamic system, which variables include at least one of a set of dynamic system parameters and a set of system outputs predictor means for generating output predictions of said multivariable dynamic system, the predictor means including; selective filtering means for selectively filtering said set of system outputs y(t); means for constructing a block state structure from the variables of said multivariable dynamic system; generating means for generating at least one of a block of k output y(t) predictions when the filtering of said set of system outputs y(t) is not performed and a block of k filtered output ψ
(t) predictions when the selective filtering of said set of system outputs y(t) is performed in accordance with the block state structure, each of dimension N consisting of output y(t) predictions and filtered output ψ
(t) predictions respectively at each future time step up to a prediction horizon k, andthe generating means including means for generating a block of k filtered output z(t) predictions using i) the block of k output y(t) predictions when the selective filtering of said set of system outputs y(t) is not performed and ii) the block of k filtered output ψ
(t) predictions when the selective filtering of said set of system outputs y(t) is performed;combining means for combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and output means for outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system. - View Dependent Claims (22, 23)
-
-
24. A block predictive controller for control and prediction of an associated multivariable dynamic system approximated as a linear system having N dynamic output signals, M dynamic input signals, and model parameter order n where N, M, and n are finite positive integers, the controller receiving a block of dynamic system outputs and generating a block of dynamic system inputs to effect said control, the controller comprising:
-
a block predictor system means adapted for receiving measurements of said dynamic system outputs, the block predictor system means including means for combining said measurements with dynamic system parameters to generate a block of predicted filtered system outputs; a block control system means for determining a block of potential system inputs based upon a block of desired output trajectory and said block of predicted filtered system outputs; and
,a control implementation unit receiving said block of potential system inputs and generating said block of dynamic system inputs according to a predetermined control algorithm to control said multivariable dynamic system. - View Dependent Claims (25)
-
-
26. An adaptive block predictive controller for control and prediction of an associated multivariable dynamic system approximated as a linear system having N dynamic output signals, M dynamic input signals, and model parameter order n where N, M, and n are finite positive integers, the controller receiving a block of dynamic system outputs and generating a block of dynamic system inputs to effect said control, the controller comprising:
-
identifier means for identifying parameters of said multivariable dynamic system and generating a block of identified parameters; a block predictor system means adapted for receiving measurements of said dynamic system outputs, the block predictor system means including means for combining said measurements with a one of dynamic system parameters from said multivariable dynamic system and said block of identified parameters to generate a block of predicted filtered system outputs; a block control system means for determining a block of potential system inputs based upon a block of desired output trajectory and said block of predicted filtered system outputs; and
,a control implementation unit receiving said block of potential system inputs and generating said block of dynamic system inputs according to a predetermined control algorithm to control said multivariable dynamic system. - View Dependent Claims (27)
-
-
28. A system output prediction method in an apparatus for predicting at least one of the filtered outputs z(t) of a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive integers, the control method comprising:
-
receiving variables including a set of dynamic system parameters and a set of system outputs from said multivariable dynamic system; generating a block of k filtered output z(t) predictions of said multivariable dynamic system by; constructing a block state structure from the variables of said multivariable dynamic system; using the block state structure, generating a block of k filtered output ψ
(t) predictions of dimension N consisting of filtered output ψ
(t) predictions at each future time step up to a prediction horizon k; and
,using a block of k filtered output ψ
(t) predictions, generating said block of k filtered output z(t) predictions.
-
-
29. A control method including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive integers, the control method comprising:
-
receiving variables including at least one of a set of dynamic system parameters and a set of system outputs y(t) from said multivariable dynamic system; generating a block of k filtered output z(t) predictions of said multivariable dynamic system by; constructing a block state structure from the variables of said multivariable dynamic system; using the block state structure, generating a block of k output y(t) predictions of dimension N consisting of output y(t) predictions at each future time step up to a prediction horizon k; and
,filtering the block of k output y(t) predictions to generate said block of k filtered output z(t) predictions; combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and
,outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system.
-
-
30. A control method including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive integers, the control method comprising:
-
receiving variables including at least one of a set of dynamic system parameters and a set of system outputs y(t) from said multivariable dynamic system; generating a block of k filtered output z(t) predictions of said multivariable dynamic system by; filtering the set of system outputs y(t); constructing a block state structure from the variables of said multivariable dynamic system; using the block state structure, generating a block of k filtered output ψ
(t) predictions of dimension N consisting of filtered output ψ
(t) predictions at each future time step up to a prediction horizon k; and
,filtering the block of k filtered output ψ
(t) predictions to generate said block of k filtered output z(t) predictions;combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and
,outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system.
-
-
31. A control method including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n where N, M, and n are finite positive integers, the control method comprising:
-
receiving variables including at least one of a set of dynamic system parameters and a set of system outputs y(t) from said multivariable dynamic system; generating a block of k filtered output z(t) predictions of said multivariable dynamic system by; filtering the set of system out,puts y(t) using a first set of filtering polynomials; constructing a block state structure from the variables of said multivariable dynamic system, the variables including the filtered set of system outputs y(t); using the block state structure, generating a block of k filtered output ψ
(t) predictions of dimension N consisting of filtered output ψ
(t) predictions at each future time step up to a prediction horizon k; and
,filtering the block of k filtered output ψ
(t) predictions to generate said block of k filtered output z(t) predictions;combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and
,outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system. - View Dependent Claims (32, 33)
-
-
34. A system controller, including system output prediction in an apparatus effecting control over a multivariable dynamic system approximated as a linear system having N output signals, M input signals, and model parameter order n whence N, M, and n are finite positive integers comprising:
-
means adapted for receiving variables from an associated multivariable dynamic system, which variables include at least one of a set of dynamic system parameters and a set of system outputs y(t); predictor means for generating output predictions of said multivariable dynamic system, the predictor means including; means for constructing a block state structure from the variables of said multivariable dynamic system; generating means for generating a block of k filtered output ψ
(t) predictions in accordance with the block state structure, the block of k filtered output ψ
(t) predictions being of dimension N and comprising filtered output ψ
(t) predictions at each future time step up to a prediction horizon k, andthe generating means including means for generating a block of k filtered output z(t) predictions using the block of k filtered output ψ
(t) predictions;combining means for combining said block of k filtered output z(t) predictions with a desired output trajectory to generate block control signals; and
,output means for outputting said block control signals as input signals to said multivariable dynamic system to effect said control over the system.
-
Specification