Fast algorithm for model predictive control
원문보기
IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0626450
(2009-11-25)
|
등록번호 |
US-8473079
(2013-06-25)
|
발명자
/ 주소 |
|
출원인 / 주소 |
- Honeywell International Inc.
|
대리인 / 주소 |
Schwegman, Lundberg & Woessner, P.A.
|
인용정보 |
피인용 횟수 :
5 인용 특허 :
1 |
초록
▼
An improved process and corresponding controller provide a model predictive control approach that can be implemented with less computational resources and/or with greater speed than conventional MPC, while at the same time retaining all or a substantial portion of the robustness and advantages of co
An improved process and corresponding controller provide a model predictive control approach that can be implemented with less computational resources and/or with greater speed than conventional MPC, while at the same time retaining all or a substantial portion of the robustness and advantages of conventional MPC. According to one aspect of the invention, the process provides an improved initial estimate of the MPC model trajectory to reduce the number of iterations to find the optimal one. The improved trajectory is obtained by applying a correction to the computed MPC manipulated value trajectory, and using the corrected manipulated value trajectory as the starting point for the next iteration of MPC manipulated value trajectory computation. As set forth in more detail below, the correction is determined from the LQR feedback control strategy. Since the sequence of control laws for the LQR feedback control strategy can be computed off-line and stored, the real time part of the LQR control strategy needed to determine the correction can be retrieved with relatively little computational resources.
대표청구항
▼
1. A method of predictive model control of a controlled system, comprising: using a model predictive control (MPC) model, determining an iterative, finite horizon optimization of a system model of the controlled system, in order to generate a manipulated value trajectory;at time t sampling a current
1. A method of predictive model control of a controlled system, comprising: using a model predictive control (MPC) model, determining an iterative, finite horizon optimization of a system model of the controlled system, in order to generate a manipulated value trajectory;at time t sampling a current state of the controlled system and computing a cost minimizing manipulated value trajectory with the MPC model for a relatively short time horizon in the future, wherein the MPC uses a quadratic programming (QP) algorithm to find the optimal solution;implementing a first step or move of the manipulated value trajectory;sampling the controlled system state again and determining the difference between the predicted and actual value of the state after the implementation of the first move;determining a correction to the manipulated value trajectory based on a first or second order approximation, using duality between the MPC model and a LQ feedback model;applying the correction to the MPC manipulated value trajectory to produce an initialized control trajectory that eliminates or reduces the need to iterate the MPC's QP algorithm to the optimal solution;repeating the control calculation starting from the now current state using the corrected manipulated value trajectory, yielding a new control and new predicted state path; andcontinuing the control process by continuing to shift the prediction horizon forward. 2. A method according to claim 1 further including using input or output blocking to reduce the computational complexity of the LQ feedback control strategy. 3. A method according to claim 1 further including computing a sequence of state feedback control laws for the LQ feedback control strategy off-line, storing the LQ control laws, and retrieving the LQ control laws for the purpose of determining the correction to the MPC manipulated value trajectory. 4. A controller used to control a controlled system, comprising: a computer system including one or more computer programs operative on the computer system to:use a model predictive control (MPC) model and determine an iterative, finite horizon optimization of a system model for the controlled system, in order to generate a manipulated value trajectory;at time t sample a current state of the controlled system and compute a cost minimizing manipulated value trajectory with the MPC model for a relatively short time horizon in the future, wherein the MPC uses a quadratic programming (QP) algorithm to find the optimal solution;implement a first step or move of the manipulated value trajectory;sample the controlled system state again and determine the difference between the predicted and actual value of the state after the implementation of the first move;determine a correction to the manipulated value trajectory based on a first or second order approximation, use duality between the MPC model and a LQ feedback model;apply the correction to the MPC manipulated value trajectory to produce an initialized control trajectory that eliminates or reduces the need to iterate the MPC's QP algorithm to the optimal solution;repeat the control calculation starting from the now current state using the corrected manipulated value trajectory, yielding a new control and new predicted state path; andcontinue the control process by continuing to shift the prediction horizon forward. 5. A system according to claim 4 further wherein the one or more computer programs use input or output blocking to reduce the computational complexity of the LQ feedback model. 6. A system according to claim 4 further wherein the one or more computer programs retrieve the LQ control laws stored for the purpose of determining the correction to the MPC manipulated value trajectory. 7. A non-transitory computer readable medium comprising instructions for executing a process of predictive model control of a controlled system, the process comprising: using a model predictive control (MPC) model, determining an iterative, finite horizon optimization of a system model of the controlled system, in order to generate a manipulated value trajectory;at time t sampling a current state of the controlled system and computing a cost minimizing manipulated value trajectory with the MPC model for a relatively short time horizon in the future, wherein the MPC uses a quadratic programming (QP) algorithm to find the optimal solution;implementing a first step or move of the manipulated value trajectory;sampling the controlled system state again and determining the difference between the predicted and actual value of the state after the implementation of the first move;determining a correction to the manipulated value trajectory based on a first or second order approximation, using duality between the MPC model and a LQ feedback model;applying the correction to the MPC manipulated value trajectory to produce an initialized control trajectory that eliminates or reduces the need to iterate the MPC's QP algorithm to the optimal solution;repeating the control calculation starting from the now current state using the corrected manipulated value trajectory, yielding a new control and new predicted state path; andcontinuing the control process by continuing to shift the prediction horizon forward. 8. The non-transitory computer readable medium of claim 7, further comprising instructions for using input or output blocking to reduce the computational complexity of the LQ feedback control strategy. 9. The non-transitory computer readable medium of claim 7, further comprising instructions for computing a sequence of state feedback control laws for the LQ feedback control strategy off-line, storing the LQ control laws, and retrieving the LQ control laws for the purpose of determining the correction to the MPC manipulated value trajectory.
이 특허에 인용된 특허 (1)
-
Thiele, Dirk, Robust adaptive model predictive controller with tuning to compensate for model mismatch.
이 특허를 인용한 특허 (5)
-
Fuxman, Adrian Matias; Pachner, Daniel, Condition-based powertrain control system.
-
Stewart, Greg; Pekar, Jaroslav; Pachner, Daniel, Coordinated engine and emissions control system.
-
Pachner, Daniel; Pekar, Jaroslav, Engine and aftertreatment optimization system.
-
Pachner, Daniel, Identification approach for internal combustion engine mean value models.
-
Markham, Thomas R., Vehicle security module system.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.