IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
UP-0999056
(2007-12-04)
|
등록번호 |
US-7826909
(2010-11-22)
|
발명자
/ 주소 |
|
인용정보 |
피인용 횟수 :
34 인용 특허 :
4 |
초록
▼
A method of dynamic model predictive control is presented in which both steady state optimization and dynamic moves calculation of the manipulated variables are determined together as part of one optimization solution (211). The method combines steady state optimization and dynamic move calculation
A method of dynamic model predictive control is presented in which both steady state optimization and dynamic moves calculation of the manipulated variables are determined together as part of one optimization solution (211). The method combines steady state optimization and dynamic move calculation such that the steady state optimal targets are determined consistent with the dynamic moves and the resulting dynamic response of the process so that the controlled variables do not violate their low/high limits in both steady state and dynamically. The method utilizes what is described as high limit dynamic violation variables and low limit dynamic violation variables corresponding to each of the controlled variables. The method offers a unique capability of mixing a part dynamic optimization with a part steady state optimization as it relates to application of a large model predictive control under varying situations, in particular at or near constraints violations that can be done dynamically in real time for improved and robust control performance without having to change the controller tuning.
대표청구항
▼
What I claim: 1. A method of operation of dynamic model predictive controller for controlling and optimizing the operation of a process having a plurality of independently controlled, manipulated variables, at least one controlled variables and none or more disturbance variables; said method of ope
What I claim: 1. A method of operation of dynamic model predictive controller for controlling and optimizing the operation of a process having a plurality of independently controlled, manipulated variables, at least one controlled variables and none or more disturbance variables; said method of operation of dynamic model predictive controller comprising the steps of: a) determining simultaneously dynamic moves of said manipulated variables and steady state values of said manipulated variables and said controlled variables as result of an optimization process combining steady state and dynamic state wherein a predetermined number of dynamic moves of said manipulated variables are determined along with steady state values of said manipulated variables and said controlled variables in accordance with steady state constraints relating to said manipulated variables and said controlled variables as well as dynamic constraints relating to said manipulated variables and said controlled variables including where appropriate relating to said disturbance variables; and b) performing a receding horizon form of control wherein said optimization is performed at successive time interval by monitoring and feedback of process responses resulting from the control actions applied at previous time intervals; wherein said optimization process further comprises an objective function J in the form of J=F (M, C, Dd, Md, Cd)+ΣΣPlc Cl+ΣΣPh cCh wherein F is some optimizing function for the process over the time horizon of time to steady state for said process, Md is dynamic values of manipulated variables over a predetermined time horizon, M is steady state value of said manipulated variables, Cd is dynamic values of controlled variables over a time horizon to steady state, C is steady state value of said controlled variables, Dd is a dynamic value vector of said disturbance variables over a time horizon no greater than said time horizon of dynamic values of said manipulated variables, and wherein further the process is considered to be a dynamic system, and said controlled variables response both in steady state and dynamic is characterized by (C, Cd)=G(Md, Dd), Phc is a penalty value to be applied for the controlled variable violating its high limit dynamically, further Chi is high limit dynamic violation variables of said controlled variables, Ci and Cli is low limit dynamic violation variables of said controlled variables, Ci, the aforementioned control variables penalty relate to either economic criteria and/or safety criteria depending on the nature and characteristics of the controlled variable. 2. The method recited in claim 1, wherein said steady state constraints comprise Ml≦M≦Mh Cl≦C≦Ch Where M is steady state value of said manipulated variables, C is steady state value of said controlled variables, Ml is low limit of said manipulated variables, M, Mh is low limit of said manipulated variables, M, Cl is low limit of said controlled variables, C, Ch is high limit of said controlled variables, C. 3. The method recited in claim 1, wherein said dynamic manipulated variables constraints comprise Ml≦Md≦Mh −ΔMjl≦ΔMj≦ΔMjh 0≦ΔMj+≦ΔMjh 0≦ΔMj−≦ΔMjl ΔMj=ΔMj+−ΔMj− −(1−kMV/k)ΔMjl≦ΔMj,k≦(1−kMVk )ΔMjh for k=1 . . . kMV where ΔMj1=Mj1−Mj* where Mj* being Current Value of Mj, ΔMj is dynamic move vector of manipulated variable, j, ΔMj+ is positive dynamic move vector of manipulated variable, j, ΔMj− is negative dynamic move vector of manipulated variable, j, ΔMjl is low limit of dynamic move of manipulated variable, j, ΔMjh is high limit of dynamic move of manipulated variable, j, ΔMjk is dynamic control move of said manipulated variable Mj at time k from now, Mj is the optimal steady state value of said manipulated variable, Mj* is the current value of said manipulated variable, j. 4. The method recited in claim 1, wherein said dynamic constraints of said controlled variables comprise Cl<=Cik<=Ch −θ≦Cik−Cik ref−Chi +Cl i≦θ 0≦Ch i 0≦Cli where Cik is predicted value of said Controlled Variable, Ci at k time interval from now, Cik ref is desired value of said Controlled Variable, Ci at k time interval from now, Chi is high limit dynamic violation variables of said Controlled Variable, Ci, Cl i is low limit dynamic violation variables of said Controlled Variable, Ci, Cik* is dynamic value of said Controlled Variable Ci at time k based on the past process condition, θ is a permitted tolerance for deviation of the predicted dynamic value of said controlled variable from its reference path, a small number, k relates to future time from now on, k=1 . . . kMV . . . kCV, kMV relates to the control horizon for manipulated variables moves, no manipulated variables to be applied beyond this time horizon so as to permit said controlled variables to attain their steady state, kCV relates to the time to steady state for said controlled variables, it would be the longest time to steady state for the changes in said manipulated variables, M plus the longest control horizon. 5. The optimizing function, F (M, C, D, Md, Cd) as recited in claim 1 can be of one of simple form such as PmM+PcC devoid of effects of dynamic values of said manipulated variables and said controlled variable on optimizing function F(), where Pm is the price value for said manipulated variables, typically a negative value representing cost and a positive value representing benefit, Pc is the price value for said controlled variables, typically a negative value-representing penalty and a positive value representing benefit. 6. Said controlled variables dynamic response, (C, Cd)=G(Md, Dd) as recited in claim 1 can be one of linear dynamic form represented as Cik=Cik*+Σgi,j(Mj−Mj*) for steady state, and as Cik=Cik*+Σgi,jkΔMjk+Σgi,lkΔDlk for dynamic response Where Ci* is the currently predicted steady state value of said controlled variable based on past changes in said manipulated variables and said disturbance variables, Ci is the steady state value of said controlled variables, Cik is predicted value of said Controlled Variable, Ci at k time interval from now, Cik* is dynamic value of said Controlled Variable Ci at time k based on the past process condition, gi,j is the steady state gain of the step response model of said Controlled Variable, Ci for a unit change in said manipulated variable, Mj, gi,jk is the step response coefficient of the process model of Controlled Variable, Ci for a unit change in said manipulated variable, Mj, ΔDlk is change in Dl at time k, gi,lk is the step response coefficient of the process model of Controlled Variable, Ci for a unit change in said disturbance variable, Dl. 7. A method of operation of dynamic model predictive controller for controlling and optimizing the operation of a process having a plurality of independently controlled, manipulated variables, at least one controlled variables and none or more disturbance variables; said method of operation of dynamic model predictive controller comprising the steps of: a) determining steady state values of said manipulated variables and said controlled variables as result of an optimization process combining steady state and dynamic state wherein with steady state values of said manipulated variables and said controlled variables in accordance with steady state constraints relating to said manipulated variables and said controlled variables and dynamic state constraints of said controlled variables including where appropriate relating to said disturbance variables; and b) performing a receding horizon form of control wherein said optimization is performed at successive time interval by monitoring and feedback of process responses resulting from the control actions applied at previous time intervals; wherein said optimization process further comprises an objective function J in the form of J=F (M, C, Dd, Cd)+ΣΣPlcCl+ΣΣPhcCh wherein F is some optimizing function for the process over the time horizon of time to steady state for the process, M is a vector of steady state value of said manipulated variables, C is a vector of steady state value of said controlled variables, Dd is a vector of disturbance variables, Cd is vector dynamic values of said controlled variables over a time horizon to steady state, Plc is a penalty value to be applied for the controlled variable violating its low limit dynamically, Phc is a penalty value to be applied for the controlled variable violating its high limit dynamically, and wherein further the process is considered to be a dynamic system, and said controlled variables dynamic response is characterized by (C, Cd)=G(M, Dd), further Chi is high limit dynamic violation variables of said controlled variables, Ci and Cli is low limit dynamic violation variables of said controlled variables, Ci. 8. The method recited in claim 7, wherein said steady state constraints comprise Ml≦M≦Mh Cl≦C≦Ch Where M is steady state value of said manipulated variables, C is steady state value of said controlled variables, Ml is low limit of said manipulated variables, M, Mh is low limit of said manipulated variables, M, Cl is low limit of said controlled variables, C, Ch is high limit of said controlled variables, C. 9. The method recited in claim 7, wherein said dynamic constraints of said controlled variables comprise C1<=Cik<=Ch −θ≦Cik−Cik ref−Chi+Cli≦θ 0≦Chi 0≦Cli where Cik is predicted value of Controlled Variable, Ci at k time interval from now, Cik ref is desired value of Controlled Variable, Ci at k time interval from now, Chi is high limit dynamic violation variables of the Controlled Variable, Ci, Cli is low limit dynamic violation variables of the Controlled Variable, Ci, Cik* is dynamic value of Controlled Variable Ci at time k based on the past process condition, θ is a permitted tolerance for deviation of the predicted dynamic value of the controlled variable from its reference path, a small number, k relates to future time from now on, k=1 . . . kMV . . . kCV, kMV relates to the control horizon for manipulated variables moves, no manipulated variables to be applied beyond this time horizon so as to permit said controlled variables to attain their steady state, kCV relates to the time to steady state for said controlled variables, it would be the longest time to steady state for the changes in said manipulated variables, M plus the longest control horizon. 10. The optimizing function, F (M, C, D, Cd) as recited in claim 7 can be of one of simple form such as PmM+PcC devoid of the effects of dynamic values of said manipulated variables and said controlled variables, where Pm is the price value for said manipulated variables, typically a negative value representing cost and a positive value representing benefit, Pc is the price value for said controlled variables, typically a negative value-representing penalty and a positive value representing benefit. 11. Said controlled variables dynamic response, (C, Cd)=G(M, Dd), as recited in claim 7 can be one of linear dynamic form represented as Ci=Ci*+Σgi,j(Mj−Mj*) for steady state, and as Cik=Cik*+Σgi,j(Mj−Mj*)+Σgi,lkΔDlk for dynamic response Where Ci* is the currently predicted steady state value of the controlled variable based on past changes in said manipulated variables and the disturbance variable, Ci is the steady state value of said controlled variables, Cik is predicted value of Controlled Variable, Ci at k time interval from now, Cik* is dynamic value of Controlled Variable Ci at time k based on the past process condition, gi,j is the steady state gain of the step response model of the Controlled Variable, Ci for a unit change in said manipulated variable, Mj, ΔDlk is change in Dl at time k, gi,ljk is the step response coefficient of the process model of Controlled Variable, Ci for a unit change in said disturbance variable, Dl. 12. A method of operation of dynamic model predictive controller for controlling and optimizing the operation of a process having a plurality of independently controlled, manipulated variables, at least one controlled variables and none or more disturbance variables; said method of operation of dynamic model predictive controller comprising the steps of: a) determining dynamic moves of said manipulated variables as result of an optimization process combining steady state and dynamic state wherein a predetermined number of dynamic moves of said manipulated variables are determined to satisfy given/known steady state values of said manipulated variables and said controlled variables in accordance with steady state constraints relating to said manipulated variables and said controlled variables as well as dynamic constraints relating to said manipulated variables and said controlled variables including where appropriate relating to said disturbance variables; and b) performing a receding horizon form of control wherein said optimization is performed at successive time interval by monitoring and feedback of process responses resulting from the control actions applied at previous time intervals; wherein said optimization process further comprises an objective function J in the form of J=F (Dd, Md, Cd)+ΣΣPlcCl+ΣΣPhcCh, wherein F is some optimizing function for the process over the time horizon of time to steady state for the process, Dd is a vector of disturbance variables, Md is vector of dynamic values of said manipulated variables over a predetermined time horizon, Cd is vector dynamic values of said controlled variables over a time horizon to steady state, Plc is a penalty value to be applied for the controlled variable violating its low limit dynamically, Phc is a penalty value to be applied for the controlled variable violating its high limit dynamically, and wherein further the process is considered to be a dynamic system, and said controlled variables dynamic response is characterized by (C, Cd)=G(Md, Dd), further Chi is high limit dynamic violation variables of said controlled variables, Ci and Cli is low limit dynamic violation variables of said controlled variables, Ci. 13. The method recited in claim 12, wherein said dynamic manipulated variables constraints comprise Ml≦Md≦Mh −ΔMjl≦ΔMj≦ΔMjh 0≦ΔMj+≦ΔMjh 0≦ΔMj−≦ΔMjl ΔMj=ΔMj+−ΔMj− ΣΔMj=Ms−M0 where ΔMj1=Mj1−Mj* where Mj* being Current Value of Mj ΔMj is dynamic move vector of manipulated variable, j, ΔMj+ is positive dynamic move vector of manipulated variable, j, ΔMj− is negative dynamic move vector of manipulated variable, j, ΔMjl is low limit of dynamic move of manipulated variable, j, ΔMjh is high limit of dynamic move of manipulated variable, j, ΔMjk is dynamic control move of said manipulated variable Mj at time k, Mj is the optimal steady state value of said manipulated variable, Mj* is the current value of said manipulated variable, j. 14. The method recited in claim 12, wherein said dynamic constraints of said controlled variables comprise Cs≦C≦Cs Cl<=Cik<=Ch −θ≦Cik−Cik ref−Chi+Cli≦θ 0≦Chi 0≦Cli where Cik is predicted value of Controlled Variable, Ci at k time interval from now, Cik ref is desired value of Controlled Variable, Ci at k time interval from now. Chi is high limit dynamic violation variables of the Controlled Variable, Ci, Cli is low limit dynamic violation variables of the Controlled Variable, Ci, Cik* is dynamic value of Controlled Variable Ci at time k based on the past process condition, Cs is steady state target of said controlled variables as given, θ is a permitted tolerance for deviation of the predicted dynamic value of the controlled variable from its reference path, a small number, k relates to future time from now on, k=1 . . . kMV . . . kCV, kMV relates to the control horizon for manipulated variables moves, no manipulated variables to be applied beyond this time horizon so as to permit said controlled variables to attain their steady state, kCV relates to the time to steady state for said controlled variables, it would be the longest time to steady state for the changes in said manipulated variables, M plus the longest control horizon. 15. The optimizing function, F (Dd, Md, Cd) as recited in claim 12 can be of one of simply null form, whereby said dynamic optimization is performed to satisfy steady state value of said manipulated variables and said controlled variables determined exogenously. 16. Said controlled variables dynamic response, Cd=G (Md, Dd) as recited in claim 12 can be one of linear dynamic form represented as Ci=Ci*+Σgi,j(Mj−Mj*) for steady state, and as Cik=Cik*+Σgi,jkΔMjk+Σgi,lkΔDlk for dynamic response Where Ci* is the currently predicted steady state value of the controlled variable based on past changes in said manipulated variables and the disturbance variable, Ci is the steady state value of said controlled variables, Cik is predicted value of Controlled Variable, Ci at k time interval from now, Cik* is dynamic value of Controlled Variable Ci at time k based on the past process condition, gi,j is the steady state gain of the step response model of the Controlled Variable, Ci for a unit change in said manipulated variable, Mj, gi,jk is the step response coefficient of the process model of Controlled Variable, Ci for a unit change in said manipulated variable, Mj, ΔDlk is change in Dl at time k, gi,lk is the step response coefficient of the process model of Controlled Variable, Ci for a unit change in said disturbance variable, Dl.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.