[TRNSYS-users] model predictive control in TRNSYS

David Bradley bradley at tess-inc.com
Wed Feb 11 12:39:59 PST 2009

There isn't any way of backing up in time in a TRNSYS simulation. I've 
given a good bit of thought to how you might do this and have never 
really figured out how to solve the problem. Fundamentally, when you 
start a simulation, there is a time when the performance is dominated to 
a smaller or larger extent by the initial values (initial temperatures 
profiles through walls, etc.). With only a 24 hour run, I would worry 
that the initial values would dominate too much.
Kind regards,

leen peeters wrote:
> Dear all,
> here is a question from a student using TRNSYS for his master's thesis:
>     ----- Original Message -----
>     *From:* Koen Bogaerts <mailto:koen.bogaerts at student.kuleuven.be>
>     *To:* Leen Peeters <mailto:Leen.Peeters at mech.kuleuven.be>
>     *Sent:* Wednesday, February 11, 2009 10:20 AM
>     *Subject:* mail trnsys
>     Dear all,
>     I'm trying to create a Model Predictive Control (MPC) for a
>     micro-CHP unit. (so a control stategy that bases it's future
>     actuator signal on what it expects to occur) 
>     Each timestep a prediction of the next 24 hours should be made.
>     After an optimization of this prediction horizon, the optimized
>     input values for that timefrane are passed to the TRNSYS model to
>     determine the CHP's working startegy for those 24 h..
>     I have my main building model in TRNSYS, i.e. the model that
>     recieves the working strategy and thus acts as the real building.
>     I wondered whether there is a way to use the same building model
>     for the prediction as well. To so so, there might be two options:
>     either: at midnight, making a copy of the TRNSYS main model (i.e.a
>     copy that has the same values for the physical state variables as
>     the main model at that timestep) and add some noise to weather
>     data and internal gains, use that copy on a larger timestep and
>     base the predictionon on that
>     either: use the main model for the prediction as well, but assume
>     you start a prediction at timestep A, run the model till timestep
>     B (24 hours later), return the simulation back to timestep A and
>     run the main simulation ,....
>     Is one of those an option in TRNSYS?
>     Regards,
>     ---
>     Koen Bogaerts
>     University of Leuven, Belgium
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm for more 
> information.
> ------------------------------------------------------------------------
> _______________________________________________
> TRNSYS-users mailing list
> TRNSYS-users at cae.wisc.edu
> https://www-old.cae.wisc.edu/mailman/listinfo/trnsys-users

Thermal Energy System Specialists (TESS), LLC 
David BRADLEY                      22 N. Carroll Street - Suite 370 
Partner                            Madison, WI 53703 
P: +1.608.274.2577 
F: +1.608.278.1475
E-mail: bradley at tess-inc.com 
Web Pages:  http://www.tess-inc.com    and      http://www.trnsys.com


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/trnsys-users-onebuilding.org/attachments/20090211/f2e9db66/attachment-0001.htm>

More information about the TRNSYS-users mailing list