Advanced AutoCSM Usage for Model Building¶
This section covers advanced topics like building custom models for Sustain-LC. It requires the following repositories and software:
Software Installation¶
Dymola and OpenModelica both provide a GUI and CLI for creating, compiling, and simulating Modelica models and exporting FMUs.
Repository Installation¶
Clone these into your working folder so that your Modelica IDE (Dymola/OpenModelica) can see them:
Modelica Buildings library
git clone https://github.com/lbl-srg/modelica-buildings.git
The Modelica Buildings Library (LBL) provides HVAC, pump, valve, and heat-exchanger models, validated against measured data for building/data-center thermal simulation.
TRANSFORM Library
git clone https://github.com/ORNL-Modelica/TRANSFORM-Library.git
Oak Ridge’s TRANSFORM toolkit models transient thermal-hydraulic systems, with detailed fluid properties and advanced reactor/cooling-loop components.
datacenterCoolingModel
git clone https://code.ornl.gov/exadigit/datacenterCoolingModel.git
ORNL’s data-center–specific CSM framework: direct-to-chip, immersion, rear-door heat-exchanger models, and facility-scale coolant loops integrated with IT thermal loads.
AutoCSM
git clone https://code.ornl.gov/exadigit/AutoCSM.git
ExaDigiT’s Python framework to auto-generate system-level cooling models: JSON→Modelica→FMU.
Custom Sustain-LC Models Using AutoCSM¶
AutoCSM reads your JSON, injects components from datacenterCoolingModel, and writes out a Modelica file. To run:
python run_auto_csm.py
Edit run_auto_csm.py to point at your JSON and set your solver parameters. The resulting FrontierModel.fmu then becomes the backend of your Gymnasium environment (frontier_env.py).