Glossary · Control theory and process control
Model predictive control (MPC)
Also known as: MPC, Receding horizon control
German: Modellprädiktive Regelung (MPC)
In control engineering, model predictive control (MPC) is a control method that uses a process model to predict future behavior over a horizon and computes the manipulated variables by solving an optimization problem with constraints at each sampling step, applying only the first move.
- Control theory
- Process industry
In one sentence
MPC predicts process behavior with a model and optimizes control moves under constraints at every step, applying only the first move.
Example
An MPC application on a polymer reactor coordinates feed rate, coolant flow and catalyst dosing to keep melt index on target without exceeding the cooling capacity.
How it applies
- Engineering: MPC handles several inputs and outputs and explicit constraints, which is why it is the workhorse of Advanced process control (APC). It relies on an identified model, typically from step tests, and often on a state estimator such as a Kalman filter.
- Operation: Operators set targets and limits rather than individual valve positions. They need to understand which constraints are active and why the controller is holding back.
- Maintenance: Model quality determines performance. Changes in equipment or feedstock call for model review; treat models as governed artifacts, similar to Model governance in systems engineering.
- Documentation: Document controlled, manipulated and disturbance variables, horizons, constraint priorities, fallback behavior and the model's validity range.
MPC vs. PID control
A PID controller reacts to the current error of one loop. MPC anticipates future behavior and coordinates several variables, at the cost of a model and more computation.