2017 08 30_meeting - lbl-srg/soep GitHub Wiki
Date: August 30, 2017, 8:15am-9am Pacific Times
The purpose of this meeting is to coordinate the Modelica OS 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.
- Kyle: Refactoring room air heat and mass balance.
- Edwin: Update about heat conduction implementation.
- Mark: OS to Modelica translation.
- Michael: Buildings API.
- Michael/Jianjun: Scalable benchmark.
- Thierry: Update about FMI-QSS.
- Kyle: Example OS measure.
- Other items
Kyle believes E+ to FMU-ME possible with ODE of room air exposed, but not easy and not his first choice.
Kyle tries to put Modelica inside E+, E+ would be main time advancing loop. Modelica would be called where E+ calls SimHVAC, Modelica would update zone nodes, but E+ does the time integration and time step control for the room air.
Kyle showed the current code, which compiles a Modelica model and in which E+ conducts the time integration of the room air temperature. (The demo is currently based on a bouncing ball problem.) Michael raised his concerns that this is a departure of the current architecture to a point where we were two to three years ago, and which we abandoned based on the reasons described in the design document of last year, such as that the coupling is across fast states and algebraic variables. Michael also said that this is the mistake many automotive companies did when adapting co-simulation and cutting the systems across domains, rather than across slow changing states. Michael said he cannot endorse this approach without having a clear design document that also addresses the mathematics, e.g., how are time steps controlled, what ODE solvers are used, how are time and state events handled, how are algebraic loops such as formed through the mass flow rate or air pressure handled. Michael also notes that Kyle's approach appears to be an approximation of QSS1, which is considerably less efficient than QSS2. Michael also said that a proof by simulation is not sufficient to use as a basis for a future design. We need to have a design, properly documented, that will build the basis for the next 10 to 20 years, that allows us using modern ODE solvers, rather than trying to invent or adapt them to get something to simulate.
For moving forward, Michael suggest that Kyle works on modifying his code so that E+ exposes an ODE as described in the current design, and if he continues his design where E+ does the time integration, it also needs to be accompanied by a design document. Michael notes that all the OS tools can still be used for either design.
Kyle said he will explore both options and we discuss on the next call, or earlier if he likes a separate meeting.
Edwin not on the call, no update.
No activity because of refocus on refactoring.
No activity.
Has been moved to master branch.
Met with Modelon to discuss FMI extension. Will meet in a week to finalize design.
For QSS, added support for input functions. Work is now on zero crossing functions that have boolean expressions.
Buildings library release scheduled for September.
None.
None.