WP1.2 Coordination Meeting March 01, 2021 - ibpsa/project1 GitHub Wiki

Meeting Report WP1.2 ‘Modelica library for MPC’

1. MEETING SUBJECT, DATE

Subject: WP1.2

Date: 01-03-2021

Location: Skype for Business

Minutes taken by: Lieve Helsen (KU Leuven)

2. PARTICIPANTS

3. AGENDA and REPORT

MARCH 1

  1. Emulators
  2. BOPTEST trials
  3. BOPTEST journal paper
  4. BS2021 BOPTEST workshop
  5. Link to ML community
  6. Suggestions for next on-line IBPSA Project 1 Expert Meeting
  7. Further issues

Emulators

SDU emulator

Tao fixed the error, basic PID controller, pushed to Github, first test needed for a full year (action Tao).

Multi-zone hydronic residential test case (Valentin)

Problems in full-year simulation, should be robust with baseline controller (Dave & Javier are looking into this), Dymola 2021 seems to work (Dave). Water boiler and radiator system has some issues with controllability due to pressure drops in the valves, switched to parametrization used in the Buildings library. Are there any rules of thumb? Some design guidelines from Michael:

  • Thermostatic valve should not have a too high pressure drop
  • High valve authority for 2-way valve
  • Non-linear valves, equal-percentage valves

BESTEST hydronic case with heat pump

  • Some updates implemented, suggested by Filip
  • Issue with event update in PyFMI: Dave implemented the fix by some extra lines of code
  • Iago starts working on his emulator again

Status:

Reporting results from BOPTEST trials

BESTEST hydronic case with heat pump:

  • Iago tested and this went well
  • Javier is testing
  • Harald did testing

BESTEST hydronic case

  • Felix is testing RL approaches
  • Harald did testing
  • Iago is prepared to test as well

BESTEST air case

  • Dave is testing

Multi-zone hydronic residential test case will be ready for testing very soon (Valentin)

Single-zone hydronic residential test case will be ready for testing very soon (Tao)

Multi-zone hydronic commercial test case is in final phase (Iago)

How can we progress faster? Any ideas?

  • We need good reporting of each controller test by May!
  • Pricing scenarios are ready
  • Functionality is implemented to choose specific time periods (still scenario to be defined that has heating and cooling). Dave sends a message as soon as this is ready to be used.
  • Data for controller model training should not be used by the KPI calculator (Action Dave)
  • Do we need extra KPIs? E.g.:
  1. operating peak demand over a month
  2. actuator travel as a measure for maintenance
  3. energy meters

Not all these values should be available for the controller (list for each emulator which measurements are available, based on practice), characterization of the controller what is needed, to decide by the user whether he/she is prepared to invest in these sensors. The measurements used are tracked and reported as a KPI (action Dave).

Outline BOPTEST journal paper focusing on methods

Authors have been contacted by Dave and outline can be found here.

In progress. 2 more weeks needed for the test case results (Javier)

BS2021 Workshop

Proposal submitted, high chance it will be approved ;-)

NeurIPS Workshop

Interaction with machine-learning community: possibility to introduce BOPTEST in December edition.

Topics for Expert meeting – breakout sessions

Michael takes action in March to plan the Expert Meeting, timing will be rather May 2021. Breakout sessions:

  1. Emulators update a. Ready b. To be completed
  2. BOPTEST trial results (incl. feedback towards workshop organizers)
  3. ...

Further issues

  1. Lieve sends a doodle to plan a next progress meeting in early April
  2. Please think all about topics for breakout sessions!
  3. Progress needed in: a. Emulators b. BOPTEST trials