Detail publikace

Generalized Predictive Control with a Non-linear Autoregressive Model

Originální název

Generalized Predictive Control with a Non-linear Autoregressive Model

Anglický název

Generalized Predictive Control with a Non-linear Autoregressive Model

Jazyk

en

Originální abstrakt

This paper presents a solution to computation of predictive control using non-linear auto-regressive models. For the non-linear model a neural network is used as a perspective tool for modelling of dynamic systems. However, the described approach is applicable to any type of auto-regressive model. The model is not linearized in the operating point, but in each control optimization step the model’s derivative is computed (linearization) for all points in the prediction horizon. The method can be used in real-time control. This is verified by porting the algorithm directly to the PLC.

Anglický abstrakt

This paper presents a solution to computation of predictive control using non-linear auto-regressive models. For the non-linear model a neural network is used as a perspective tool for modelling of dynamic systems. However, the described approach is applicable to any type of auto-regressive model. The model is not linearized in the operating point, but in each control optimization step the model’s derivative is computed (linearization) for all points in the prediction horizon. The method can be used in real-time control. This is verified by porting the algorithm directly to the PLC.

BibTex


@article{BUT46302,
  author="Hynek {Vychodil} and Michal {Schmidt} and Petr {Nepevný} and Petr {Pivoňka}",
  title="Generalized Predictive Control with a Non-linear Autoregressive Model",
  annote="This paper presents a solution to computation of predictive control using non-linear auto-regressive
models. For the non-linear model a neural network is used as a perspective tool for modelling of dynamic systems.
However, the described approach is applicable to any type of auto-regressive model. The model is not linearized
in the operating point, but in each control optimization step the model’s derivative is computed (linearization)
for all points in the prediction horizon. The method can be used in real-time control. This is verified by porting
the algorithm directly to the PLC.",
  chapter="46302",
  journal="WSEAS Transactions on Circuits",
  number="3",
  volume="2005",
  year="2005",
  month="march",
  pages="125",
  type="journal article - other"
}