Stepwise introduction of MPC in industrial control systems: a PID–MPC hybrid
(2026)Department of Automatic Control
- Abstract
- Industrial control systems must ensure reliable operation while respecting physical and operational constraints. In practice, PID controllers dominate industrial control, often cited as covering over 95% of control loops. PID controllers are reliable, well understood, and easy to maintain, but they cannot explicitly handle constraints. In contrast, model predictive control (MPC) can enforce constraints directly, yet replacing an existing PID loop with MPC is often high-risk due to modeling and tuning requirements, reduced operator familiarity, and disruption in production.
This thesis extends, implements, and evaluates a seamless PID–MPC hybrid architecture, where the existing control structure is retained and an MPC layer computes an... (More) - Industrial control systems must ensure reliable operation while respecting physical and operational constraints. In practice, PID controllers dominate industrial control, often cited as covering over 95% of control loops. PID controllers are reliable, well understood, and easy to maintain, but they cannot explicitly handle constraints. In contrast, model predictive control (MPC) can enforce constraints directly, yet replacing an existing PID loop with MPC is often high-risk due to modeling and tuning requirements, reduced operator familiarity, and disruption in production.
This thesis extends, implements, and evaluates a seamless PID–MPC hybrid architecture, where the existing control structure is retained and an MPC layer computes an additive correction w to the PID output v. The MPC is formulated on the plant and PID closed-loop model and solved as a quadratic program suitable for real-time execution. The architecture allows for a stepwise introduction of MPC control through a scaling parameter α.
The formulation enforces constraints while maintaining feasibility under disturbances through soft constraints. Offset-free tracking is achieved via disturbancestate augmentation and a Kalman filter, where inclusion of the measured PID output improves observability. A smoothing parameter β regularizes the correction signal and is found to be crucial for robust behavior in the presence of noise and implementation effects.
The approach is validated in simulation on benchmark processes and in realtime experiments. Results show improved constraint handling compared to a PIDonly baseline and demonstrate that MPC functionality can be introduced stepwise via α. The work also identifies the phenomenon of constraint-induced integrator windup, arising from the interaction between control signal constraints and integral action in the PID loop. Experimental validation on an ABB AC 800M controller connected to a physical quadruple-tank process demonstrates practical feasibility beyond simulation. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9242891
- author
- Månsson, Adam and Gustafsson, Simon
- supervisor
- organization
- year
- 2026
- type
- H3 - Professional qualifications (4 Years - )
- subject
- keywords
- PID–MPC, PID, MPC, Hybrid control, Industrial Control System, Constraint handling, ABB System 800xA, Quadratic Programming, Quadruple-tank
- report number
- TFRT-6308
- other publication id
- 0280-5316
- language
- English
- id
- 9242891
- date added to LUP
- 2026-06-29 14:39:38
- date last changed
- 2026-06-29 14:39:38
@misc{9242891,
abstract = {{Industrial control systems must ensure reliable operation while respecting physical and operational constraints. In practice, PID controllers dominate industrial control, often cited as covering over 95% of control loops. PID controllers are reliable, well understood, and easy to maintain, but they cannot explicitly handle constraints. In contrast, model predictive control (MPC) can enforce constraints directly, yet replacing an existing PID loop with MPC is often high-risk due to modeling and tuning requirements, reduced operator familiarity, and disruption in production.
This thesis extends, implements, and evaluates a seamless PID–MPC hybrid architecture, where the existing control structure is retained and an MPC layer computes an additive correction w to the PID output v. The MPC is formulated on the plant and PID closed-loop model and solved as a quadratic program suitable for real-time execution. The architecture allows for a stepwise introduction of MPC control through a scaling parameter α.
The formulation enforces constraints while maintaining feasibility under disturbances through soft constraints. Offset-free tracking is achieved via disturbancestate augmentation and a Kalman filter, where inclusion of the measured PID output improves observability. A smoothing parameter β regularizes the correction signal and is found to be crucial for robust behavior in the presence of noise and implementation effects.
The approach is validated in simulation on benchmark processes and in realtime experiments. Results show improved constraint handling compared to a PIDonly baseline and demonstrate that MPC functionality can be introduced stepwise via α. The work also identifies the phenomenon of constraint-induced integrator windup, arising from the interaction between control signal constraints and integral action in the PID loop. Experimental validation on an ABB AC 800M controller connected to a physical quadruple-tank process demonstrates practical feasibility beyond simulation.}},
author = {{Månsson, Adam and Gustafsson, Simon}},
language = {{eng}},
note = {{Student Paper}},
title = {{Stepwise introduction of MPC in industrial control systems: a PID–MPC hybrid}},
year = {{2026}},
}