[TRNSYS-users] Type 22

Michaël Kummert michael.kummert at gmail.com
Thu Jul 21 09:07:33 PDT 2005


Krishna,

Type 22 is best adapted to "almost-algebraic" equations, i.e. to systems 
where the thermal mass is low enough so that the time constant is not 
large compared to the time step. This is the case for example when you 
use Type1 solar collector or other components that implement 
steady-state equations.
Type 22 won't work when you have a long delay between the response of 
the system and a change in the control signal (where "long" should be 
defined in comparison with the chosen time step), or when the "slope" of 
the relationship between a change in control signal and the resulting 
change in the control variable is almost zero. Flowrate control in a 
floor heating is a good example of something that would not work because 
of the time delay: a large change in flowrate ususally results in a very 
small change in zone temperature at the same time step, so the line 
search algorithm is useless or at best it ends up acting as a bang-bang 
controller.

As Dave mentioned, using a component that delays its inputs (e.g. Type 
93) would help convergence. However, it would require using a relatively 
short time step, in which case you can also use Type 23, the PID 
controller, in "real-time" mode (the default mode). In that mode Type 23 
  gets the converged values of one time step and calculates the control 
signal for the next time step (which remains constant through the 
iterations). By tuning the PID you will be able to get some anticipation 
in your controller, which you won't have using Type 93 + Type 22 or 
another line search algorithm.

Very often convergence problems with controllers come from people 
wanting to mimic actual controllers with a large time step (e.g. 1 hour) 
while those actual controllers use a much shorter time step and don't 
iterate. Most of the time I find that the simulation is actually more 
efficient when you decrease the time step and use "real-time" 
controllers (i.e. controllers that don't iterate): you have more time 
steps but less iterations per time steps.

For the On/Off control of auxiliary equipment, I would use an On/Off 
controller based on temperature and also have that controller use the 
temperature at previous time step (assuming you have a short time step).

I hope this helps,

Michaël Kummert

_________________________________________________________________

Michaël Kummert

École Polytechnique de Montréal - Génie Mécanique
Case Postale 6079, succursale Centre-Ville
Montréal   QC   H3C 3A7
Canada

Tel: +1 (514) 340-4711, Ext. 3367
Fax: +1 (514) 340-5917
Email : michael.kummert à polymtl.ca



More information about the TRNSYS-users mailing list