Experimental and Simulation Investigation of an Adaptive Model Predictive Control Scheme: Model Parametrized by...

Area of Science:

• Process Control
• Adaptive Systems
• Chemical Engineering

Background:

• Model predictive control (MPC) performance depends on accurate prediction models.
• Linear MPC models, common in continuous plants, falter during significant operating point shifts.
• Maintaining MPC effectiveness requires updating the prediction model for changing nominal conditions.

Purpose of the Study:

• Design an adaptive model predictive control (AMPC) scheme.
• Utilize linear models estimated from input-output perturbation data.
• Address limitations of fixed-model MPC in dynamic industrial environments.

Main Methods:

• Developed an adaptive MPC (AMPC) scheme using OBF-ARX (generalized orthonormal basis filters with ARX structure) for observer dynamics.
• Estimated linear models from perturbation data under nominal operating conditions.
• Proposed both fixed and variable pole AMPC schemes.

Main Results:

• Demonstrated AMPC efficacy via simulations on a binary distillation column.
• Validated AMPC performance through experimental studies on a two-tank heater setup.
• Confirmed effectiveness in both servo (set-point tracking) and regulator (disturbance rejection) problems.

Conclusions:

• The proposed AMPC schemes effectively maintain control performance under changing operating conditions.
• AMPC successfully tracks set points and rejects disturbances in dynamic systems.
• These adaptive model predictive control strategies show significant potential for precise industrial process control.