Model predictive controllers use linear models to control both linear and nonlinear plants that run within a local operating range. Plants with complex characteristics such as long time delays, higher-order dynamics, or strong interactions are particularly well-suited for model predictive control. To create a plant model, you can directly specify a linear model, linearize a Simulink® model, or identify a linear model using measured data. When creating a plant model for use in model predictive control, it is important to specify the input and output signal types and scale factors. For more information, see MPC Signal Types and Specify Scale Factors.
Model and Signals
- MPC Signal Types
Plant inputs are independent variables that affect the plant, and plant outputs are dependent variables that you want to control or monitor.
- MPC Prediction Models
Model predictive controllers use plant, disturbance, and noise models for prediction and state estimation.
- Specify Scale Factors
When designing an MPC controller, it is good practice to define scale factors for each plant input and output, especially when variables have large differences in magnitude.
Obtain LTI Models
- Construct Linear Time Invariant Models
MPC controllers support the same LTI model formats as Control System Toolbox™ software.
- Specify Multi-Input Multi-Output Plants
Most MPC applications involve plants with multiple inputs and outputs.
- Linearize Simulink Models
Obtain a linear approximation of a nonlinear plant at a specified operating point.
- Linearize Simulink Models Using MPC Designer
Open MPC Designer from Simulink and define the MPC structure by linearizing the model.
- Identify Plant from Data
Estimate a linear System Identification Toolbox™ model using measured input/output data.
- CSTR Model
Description of a continuously stirred-tank reactor (CSTR) involving an exothermic reaction.