[TRNSYS-users] How to speed-up my co-simulation?

[Amirreza Heidari]

Hello,

I need to run my TRNSYS simulation from a Python script to give some inputs
and then read back some outputs. At the moment I am using I write values to
the .back  and .b18 files, then the *subprocess *function in Python runs
the simulation for one timestep, and then I read the outputs from the
outputs of printer component. The co-simulation with this method is very
time-consuming. So I am searching for easier methods. I also tried the
following:

*Implementing my Python script in Type 3157**:* This component has a nested
dictionary to receive inputs and give back outputs. But my Python script
includes a Neural Network that should be trained, and it is not clear how
this component can save the neural network. Saving it externally is not
feasible due to some other details.

*Using FMU*: I heard about TRNSYS FMU
<https://sourceforge.net/projects/trnsys-fmu/> for co-simulation but it
does not work with my 64-bit system.

*Simplifying the simulation*: I removed all the output types and other
unnecessary types.

Any feedback on how to speed up my co-simulation is appreciated.

Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/trnsys-users-onebuilding.org/attachments/20230202/5b3e1a27/attachment.htm>