-
Notifications
You must be signed in to change notification settings - Fork 3
2019 09 11_meeting
Date: September 11, 2019, 8:15am-9am Pacific Times
The purpose of this meeting is to coordinate the Modelica integration for SOEP.
Join from PC, Mac, Linux, iOS or Android: https://lbnl.zoom.us/my/mwetter
Or iPhone one-tap (US Toll): +14086380968,6614042296# or +16465588656,6614042296#
- Or Telephone:
- Dial: +1 408 638 0968 (US Toll) or +1 646 558 8656 (US Toll) Meeting ID: 661 404 2296 International numbers available: https://lbnl.zoom.us/zoomconference?m=_h5BuJ686mPy3rWEyKf4NROxLaeOV5J8
- Previous action items.
- Modelon update about numerical investigation (Agnes).
- Kyle, Michael: Refactoring room air heat and mass balance.
- EnergyPlus I/O variables.
- Stuart: FMI-QSS.
- Kyle, Michael: Optimica license protection.
- Jianjun: JSON representation for OS front-end to Modelica
- Other items
Agnes presented the analysis of the nonlinear equations.
Jianjun will change nle_solver_tol_factor to 1E-2 instead of 1E-4 in regression testing.
Michael will look into making dp of pressure drop elements a state rather than p using stateSelect.
Fall back to avoid tearing should be added to next subcontract.
Agnes to get back regarding debug information for user for nonlinear solver.
No discussion. Will need updated EnergyPlus FMU from Kyle.
Branch issue1129_energyPlus_zone now works for one building for JModelica, Dymola and OpenModelica.
Branch issue1129_energyPlus_zone_fmiSignature works with one and multiple buildings exported from JModelica as an FMU.
Next, work on
- Kyle to look into numerical instability (walls may be initialized to wrong temperature)
- Michael to look into where segfault happens for jm_ipython.sh jmodelica.py Buildings.Experimental.EnergyPlus.Validation.ThreeZonesTwoBuildings It works for JModelica generated FMUs, but not for EnergyPlus. However, Kyle has a unit test that works for multiple EnergyPlus FMUs.
- Kyle to add error handling and flag to clean temporary files.
- Kyle to provide an idf file with 3 identical rooms.
- Installer package for all three platforms.
- Then implement control I/O.
- All zones in the idf file must have a zone model in Modelica. Otherwise the simulation stops with an error.
- If there is an HVAC system in the idf file, then EnergyPlus issues a warning, the EnergyPlus HVAC system is not simulated, but the coupled EnergyPlus/Modelica simulation proceeds.
- For the EnergyPlus envelope, either the CTF transfer function or the finite difference method can be used.
No activity until room model is finished.
No discussion.
Stuart works on using the directional derivatives, has it working now in OCT. Now updates QSS code to use directional derivatives, sees no significant performance hit.
Stuart works with Agnes as reverse dependencies seem missing in the modelDescription.xml file.
No discussion.
Next:
- Test using funnel as the default diff tool in BuildingsPy.
- Use funnel for cross comparison between dymola and jmodelica.
- Work on experiment specification.
Subcontract between Modelon and NREL still not in place. Hubertus will have a call with Luigi.
Potential paths: Check diff to frozen library. Allow n models to be changed. Upgrade path to full OCT version (with access to cloud version in an integrated way).
Need to define the different use cases and work flows.
Use of python API makes protection much easier, is sufficient unless interactive mode is needed. Java API would be needed for interactive mode.
No discussion.
Jianjun to work on modelica-json parsing in Modelica rather than CDL mode.
Ongoing is dealing with replaceable models.
None.
Jianjun will change nle_solver_tol_factor to 1E-2 instead of 1E-4 in regression testing.
Michael will look into making dp of pressure drop elements a state rather than p using stateSelect.
Michael to add to next Modelon subcontract fall back to avoid tearing.
Agnes to get back regarding debug information for user for nonlinear solver.
Kyle to produce updated EnergyPlus FMU that does not show large convective load (which may be caused by wrong initialization).
None.