Difference between revisions of "Dp 364 en"

From DCEwiki
Jump to: navigation, search
Line 4: Line 4:
 
   
 
   
 
[[Diplomové práce 2009]]
 
[[Diplomové práce 2009]]
 +
 +
[[Image:364_dp_en.gif]]
  
 
Model Predictive Control (MPC) is a popular method which can naturally deal with
 
Model Predictive Control (MPC) is a popular method which can naturally deal with
Line 13: Line 15:
 
of MPC is complexity of the result control law, which increases with the number of
 
of MPC is complexity of the result control law, which increases with the number of
 
constraints.
 
constraints.
 +
 
The weakness of MPC is the need for a good and accurate process model because the
 
The weakness of MPC is the need for a good and accurate process model because the
 
model is used for prediction of future system response. When disturbances or model
 
model is used for prediction of future system response. When disturbances or model
Line 24: Line 27:
 
despite of the small complexity of the control law, presence of model uncertainty and
 
despite of the small complexity of the control law, presence of model uncertainty and
 
disturbances can be solved by using the algorithm described in this work.
 
disturbances can be solved by using the algorithm described in this work.
 +
 +
* '''Šantin Ondřej''', mailto:ondrej.santin@gmail.com
 +
* '''Pekař Jaroslav''', mailto:Jaroslav.Pekar@Honeywell.com

Revision as of 16:58, 5 May 2010

Influence of Model Uncertainty on Constraints Handling in Predictive Control

Author: Šantin Ondřej

Diplomové práce 2009

364 dp en.gif

Model Predictive Control (MPC) is a popular method which can naturally deal with constraints and provides optimal control actions based on a declared cost function. The main drawback of this approach is need for on-line optimization, which limits the usage of MPC only for slow process. The on-line optimization can be avoided by using the Explicit formulation of MPC, where all optimization can be performed off-line, so that it is possible to use MPC even for high-speed systems. Main drawback of Explicit formulation of MPC is complexity of the result control law, which increases with the number of constraints.

The weakness of MPC is the need for a good and accurate process model because the model is used for prediction of future system response. When disturbances or model uncertainty are present, the control performance may be poor and the solution of optimization may be infeasible in constrained case. Thus many practical applications use the range control, where the controlled variables are enabled to freely vary in specified range and violation of the range is penalized by a quadratic function. The problem is that the range control needs to add many constraints so that the complexity of resulting explicit MPC control law will extremely increase. The conflict that it is necessary to ensure the feasible solution of the constrained MPC, despite of the small complexity of the control law, presence of model uncertainty and disturbances can be solved by using the algorithm described in this work.