System and method for implementing model predictive control in PLC
View Patent ↗A Model Predictive Control (MPC) framework is implemented as part of the run-time system function features of a Programmable Logic Controller (PLC) system. Optimal control calculations are performed in the run-time MPC function block of the PLC. The optimal control function is determined by an MPC block in an engineering tool of the PLC, using a system dynamic matrix containing measurements from a unit step response test performed by the PLC.
1. A method for controlling process equipment via a programmable logic controller, comprising:
by a processor implementing a model predictive control engineering tool, receiving model predictive control engineering parameters;
by the processor, commanding the programmable logic controller to perform a model identification operation on the process equipment;
by the processor, receiving, from the programmable logic controller, results from the model identification operation;
by the processor, constructing a system process model for the process equipment, using the results from the model identification operation;
by the processor, implementing a programmable logic control engineering tool to construct a control logic for the process equipment; and
downloading the system process model to a model predictive control function block in the programmable logic controller and the control logic to a proportional/integral/derivative function block in the programmable logic controller, both function blocks called to implement process equipment control.
2. A method as in claim 1 , wherein the model predictive control engineering parameters comprise: an error correction factor array, a prediction horizon and a cost weight coefficient array.
3. A method as in claim 1 , wherein the system process model is a dynamic matrix model.
4. A method as in claim 3 , wherein the model identification operation comprises a unit step open loop response test.
5. A method as in claim 4 , wherein the unit step open loop response test comprises:
transmitting a control command to the process equipment via the programmable logic controller; and
receiving by the programmable logic controller a series of response measurements from the process equipment.
6. A method as in claim 5 , wherein the transmitting is through an output module of the programmable logic controller and the receiving is through an input module of the programmable logic controller.
7. A method as in claim 3 , wherein the system process model implements the following:
Δ U ( k )=( A T QA +λ) −1 A T Q[Y r ( k+ 1)− A 0 ΔU ( k− 1)− he ( k )]
where ΔU(k) is an unknown control increment vector, A and A 0 comprise the system dynamic matrix, Q and λ are cost weighting functions, Y r (k+1) is a desired output for a future P instants, ΔU(k−1) is a known control increment vector, h is an error correction faction and e(k) is an output error of the model at time k.
8. A method as in claim 1 , wherein the system process model is a model selected from the group consisting of a state-space representation, an autoregressive model with external input and an autoregressive moving average model with external input.
9. A method as in claim 1 , wherein the model predictive control engineering tool and the programmable logic control engineering tool comprise a tool suite with a single human-machine interface.
10. A non-transitory computer-usable medium having computer readable instructions stored thereon that, when executed by a processor, cause the processor to perform operations for controlling process equipment via a programmable logic controller, the operations comprising:
receiving model predictive control engineering parameters;
commanding the programmable logic controller to perform a model identification operation on the process equipment;
receiving, from the programmable logic controller, results from the model identification operation;
constructing a system process model for the process equipment, using the results from the model identification operation in a model predictive engineering tool;
constructing a control logic for the process equipment in a programmable logic control engineering tool; and
downloading the system process model to a model predictive control function block in the programmable logic controller and the control logic to a proportional/integral/derivative function block in the programmable logic controller, both function blocks called to implement process equipment control.
11. A non-transitory computer-usable medium as in claim 10 , wherein the model predictive control engineering parameters comprise: an error correction factor array, a prediction horizon and a cost weighting coefficient array.
12. A non-transitory computer-usable medium as in claim 10 , wherein the system process model is a dynamic matrix model.
13. A non-transitory computer-usable medium as in claim 12 , wherein the model identification operation comprises a unit step open loop response test.
14. A non-transitory computer-usable medium as in claim 13 , wherein the unit step open loop response test comprises:
transmitting a control command to the process equipment via the programmable logic controller; and
receiving by the programmable logic controller a series of response measurements from the process equipment.
15. A non-transitory computer-usable medium as in claim 14 , wherein the transmitting is through an output module of the programmable logic controller and the receiving is through an input module of the programmable logic controller.
16. A non-transitory computer-usable medium as in claim 12 , wherein the system process model implements the following:
Δ U ( k )=( A T QA +λ) −1 A T Q[Y r ( k+ 1)− A 0 ΔU ( k− 1)− he ( k )]
where ΔU(k) is an unknown control increment vector, A and A 0 comprise the system dynamic matrix, Q and λ are cost weighting functions, Y r (k+1) is a desired output for a future P instants, ΔU(k−1) is a known control increment vector, h is an error correction faction and e(k) is an output error of the model at time k.
17. A non-transitory computer-usable medium as in claim 10 , wherein the system process model is a model selected from the group consisting of a state-space representation, an autoregressive model with external input and an autoregressive moving average model with external input.
18. A non-transitory computer-usable medium as in claim 10 , wherein the model predictive control engineering tool and the programmable logic control engineering tool comprise a tool suite with a single human-machine interface.
19. A model predictive control engineering tool for programming a programmable logic controller to implement model predictive control, the engineering tool comprising a processor and a non-transitory computer-usable medium having computer readable instructions stored thereon that, when executed by the processor, cause the processor to perform operations for controlling process equipment via a programmable logic controller, the operations comprising:
receiving model predictive control engineering parameters;
commanding the programmable logic controller to perform a model identification operation on the process equipment;
receiving, from the programmable logic controller, results from the model identification operation;
constructing a system process model for the process equipment, using the results from the model identification operation;
constructing a control logic for the process equipment utilizing an included programmable logic control engineering tool; and
downloading the system process model to a model predictive control function block in the programmable logic controller and the control logic to a proportional/integral/derivative function block in the programmable logic controller, both function blocks called to implement process equipment control.
20. A model predictive control engineering tool as in claim 19 , wherein the system process model is a dynamic matrix model.