Predictive control: modeling - Predictive control : Process Modelling

This article is aimed at technicians and engineers responsible for installing automatic control systems on industrial production sites. The Predictive Controller proposed here is being used more and more, as it has features that respond well to users' needs, with an attractive cost/performance ratio. With the aim of making the message easily accessible, this control is presented here in the simplest possible way, while mathematical justifications are available in other more theoretical documents.

It takes the form of a control algorithm that can be implemented, or often already pre-implemented, in the PLCs and control systems currently available on the industrial market.

All processes in their own environments are unique, with their own specific objectives. The aim is not to apply "recipes", but rather to provide a basis for solutions which the local automation specialist can use to adapt the method and its settings to the specific case.

The control law uses a real-time dynamic process model, the identification of which is the main task.

PFC control is the oldest form of predictive control, and was deliberately designed to be "easy to understand" and implement, but it generally delivers better performance in terms of accuracy and stability than traditional PID control, which it does not seek to replace, but rather to take over when it reaches its limits.

Predictive control is fundamentally model-based, and so we begin by describing the problems and proposed solutions for obtaining these models, which will be implemented and used in real time in the controller.