A method controls an operation of a system. The system is operated by a controller. The controller is model-based controller determined according to a model of the system. The method updates the model, during the operation, based on an extremum seeking, and updates the controller based on the update
A method controls an operation of a system. The system is operated by a controller. The controller is model-based controller determined according to a model of the system. The method updates the model, during the operation, based on an extremum seeking, and updates the controller based on the updated model.
대표청구항▼
1. A controller for controlling an operation of a system, comprising: a processor for executing a learning module, wherein the learning module updates a model of the system using an extremum seeking, wherein the learning moduledetermines a value Q[k] of a cost function at an iteration k, according t
1. A controller for controlling an operation of a system, comprising: a processor for executing a learning module, wherein the learning module updates a model of the system using an extremum seeking, wherein the learning moduledetermines a value Q[k] of a cost function at an iteration k, according to Q[k]=C1({dot over (x)}ref[k]−{dot over (x)}est[k])2+C2(xref[k]−xest[k])2, wherein C1, and C2 are positive constants, xref is a reference state of the system and xest is an actual state of the system; and determines a coefficient βi of the model according to yi[k+1]=yi[k]+aiΔT sin(ωikΔT+φ)Q[k]βi[k]=yi[k]+ai sin(ωikΔT+φ), for i=1, . . . n, wherein yi[k] is the value of an intermediate integration variable at the iteration k, ΔT is a sampling time, i.e. t=kΔT, ai sin(ωikΔT+φ) is a perturbation signal, wherein the perturbation signal is a sinusoid function with an amplitude ai, a frequency ωi, and a phase φ, i=1, . . . n, n is a number of perturbation signals and βi[k] is a value of the coefficient βi at the iteration k. 2. The controller of claim 1, wherein the learning module minimizes a cost function representing an error between an actual and a desired output of the system; andupdates the model based on a value of the minimized cost function and a perturbation signal. 3. The controller of claim 1, wherein the learning module updates the model iteratively during the operation of the system. 4. A method for controlling an operation of a system, comprising: determining a value Q[k] of a cost function at an iteration k, according to Q[k]=C1({dot over (x)}ref[k]−{dot over (x)}est[k])2+C2(xref[k]−xest[k])2, wherein C1, and C2 are positive constants, xref is a reference state of the system and xest is an actual state of the system; determining a coefficient βi of the model according to yi[k+1]=yi[k]+aiΔT sin(ωikΔT+φ)Q [k]βi[k]=yi[k]+ai sin(ωikΔT+φ), for i=1, . . . n, wherein yi[k] is the value of an intermediate integration variable at the iteration k, ΔT is a sampling time, i.e. t=kΔT, ai sin(ωikΔT+φ) is a perturbation signal, wherein the perturbation signal is a sinusoid function with an amplitude ai, a frequency ωi, and a phase φ, i=1, . . . n, n is a number of perturbation signals and βi[k] is a value of the coefficient βi at the iteration k; and updating the coefficient βi the model of the system, wherein steps of the method are performed by a processor. 5. The method of claim 4, further comprising: updating a controller operating the system based on the updated model. 6. The method of claim 5, further comprising: updating the model and the controller iteratively during the operation of the system. 7. A method for controlling an operation of a system, comprising: operating the system by a controller, wherein the controller is model-based controller determined according to a model of the system;updating the model, during the operation, based on an extremum seeking, comprising: determining a value Q[k] of a cost function at an iteration k, according to Q[k]=C1({dot over (x)}ref [k]−{dot over (x)}est[k])2+C2(xref[k]−xest[k])2, wherein C1, and C2 are positive constants, xref is a reference state of the system and xest is an actual state of the system; and determining the coefficient βi the model according to yi[k+1]=yi[k]+aiΔT sin(ωikΔT+φ)Q[k]βi[k]=yi[k]+ai sin(ωikΔT+φ), for i=1, . . . n, wherein yi[k] is the value of an intermediate integration variable at the iteration k, ΔT is a sampling time, i.e. t=kΔT, ai sin(ωikΔT+φ) is a perturbation signal, wherein the perturbation signal is a sinusoid function with an amplitude ai, a frequency ωi, and a phase φ, i=1, . . . n, n is a number of perturbation signals and βi[k] is a value of the coefficient βi at the iteration k; and updating the controller based on the updated model, wherein steps of the method are performed by a processor. 8. The method of claim 7, wherein the system is a break system including an actuator having a moving element, and wherein the reference state is a reference position of the moving element, and the actual state is an actual position of the moving element.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (2)
Dong Yang (Cleveland Heights OH) Chizeck Howard J. (Cleveland Heights OH) Khoury James M. (Strongsville OH) Schmidt Robert N. (Cleveland OH), Extended horizon adaptive block predictive controller with an efficient prediction system.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.