-
Notifications
You must be signed in to change notification settings - Fork 3
2017 01 18_meeting
Date: Jan. 18, 2017, 8am-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/j/6614042296
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
- David: Setup for "minimum viable product" with an end-to-end demonstration for a 1 zone model and an idealized system.
- Michael: Update about CTF implementation (code stub for Edwin to work on).
- Michael: Scalable benchmark (about to start working on it).
- Michael: Update about FMI-QSS.
- DOE peer-review: Slides are due to DOE by Feb. 27.
Dave showed slides with first steps in model for minimum viable product. We will first focus on a single zone VAV system, and later in the fiscal year also use a 5 zone VAV system that ORNL started building. Mark said that the model is close to be ready. Michael prefers to start with 1 zone model to keep it simple, to faster debug in case of differences between E+ and Modelica, and also because work on QSS and the heat conduction needs to progress before we can look at computing performance. Mark said that some OpenStudio design work requires multiple zone. Michael said we could run them in parallel, but focus the translation first on a simple 1 zone model to make sure we get this right before adding complexity and increasing computing time. Mark said that for the next FY, we need larger models for users. Michael agrees.
Kyle said he will work on an OS measure to see how to 'pre-package' higher level components. We will discuss this on the next call with the intent to iterate on what information should be put in OS vs. Modelica vs. generated from Modelica. Michael said that there are two aspect: (i) what parameters and variables to expose to the user (and with what name), and (ii) what prepackaged system model (such as a VAV system for N zones and replaceable components) to expose to the user.
Michael showed the Modelica and C stub for the CTF implementation. Edwin started working on an implementation in a separate github repository. He said that in CTF, taking Q_flow as an input and computing T would require an iteration because CTF are set up to take T as an input and output Q_flow. Michael prefers to use T as an output, as otherwise, the radiation network will need to be solved iteratively for the temperature (involving all surfaces). Edwin will look into using the Periodic Response Factor method which may be better applicable than CTF because it outputs T.
Before implementation, Edwin will estimate the number of operations (multiplications and additions) for the Periodic Response Factor method and compare it against a finite difference method to see how much performance improvement is realistic.
LBNL will start setting up a model with N rooms that will be used to test how the computations scale. Currently, there is no such test case in E+.
Work is progressing on implementing QSS. Next is to see how to package a Modelica generated FMU as an FMU that contains the QSS solver.
- Michael to work on small prototypical building in Modelica.
- Kyle to work on a measure for minimum viable product.
- Mark to rescope ORNL FY 17 work.
- Dave to add VAV model to 1 zone test case.
- Kyle to prepare OS measures in order to move forward with how to design OS-Modelica API.