Model Based Control of Refrigeration Containers

1
Model Based Control of Refrigeration Containers

• Kresten Kjær Sørensen
• Lodam electronics a/s and
• Aalborg University

2

• Outline
• Developmental, commercial and scientific
  objectives
• System overview
• Project status
• Issues that should be addressed
• Research

3

• Developmental Objectives
• A set of methods that enable optimal,
  adaptive, constrained control of a refrigeration
  container that is also applicable to other
  thermodynamical systems

4

• Commercial Objectives
• Better competitiveness through
• Reduced energy consumption
• Reduced mechanical wear and tear
• Improved adaptiveness of the controller

5

• Scientific Objectives
• To use MPC for refrigeration by
• Finding a suitable method for modeling nonlinear
  dynamics of the refrigeration system that may be
  used by an MPC.
• Finding a method to cope with changes in system
  dynamics in the event of failure.

6

• System overview

• Known refrigeration system dynamics
• Unknown cargo dynamics
• Known constraints on cargo temperature
• Economizer gives an additional degree of freedom

7

• Log(p)-h Diagram

8

• Project Status
• A library of nonlinear refrigeration component
  models has been developed.
• A composite simulation model has been developed.
• A framework that allows to test a controller,
  working within Matlab, with real hardware is
  ready.

9

• Simulation Model 1/2
• Nonlinear simulation model of the refrigeration
  system.
• Composed of smaller component models from a
  library.
• Simulation is done in Matlab by a set of custom
  functions.

10

• Simulation Model 2/2

• The simulation model is described by a number of
  files
• Each component are simulated separately for a
  short time step
• All outputs at the end of the simulation is
  propagated to the relevant inputs
• Simulation run again for another time step
• Equivalent of simulating a large, combined
  function with ZOHs inserted between components

11

• Issues that must be addressed
• The system is nonlinear and has a large area of
  operation for both hot and cold temperatures.
• There are discrete actuators with strong
  influence on the dynamics that cannot be
  neglected.
• In case of partial system malfunction it is very
  important to continue cooling.

12

• Research
• Adaptive MPC for increased robustness in the
  event of failure.
• Auto-generation of a model usable with MPC, based
  on the simulation model.