WP1.2 Coordination Meeting October 8, 2019

Meeting Report WP1.2 ‘Modelica library for MPC’

1. MEETING SUBJECT, DATE

Subject: WP1.2

Date: 08-10-2019

Location: Skype for Business

Minutes taken by: Lieve Helsen (KU Leuven)

2. PARTICIPANTS

Company/Organisation Participant

KU Leuven Lieve Helsen (WP Leader)

KU Leuven Filip Jorissen

KU Leuven Iago Cupeiro (now @ Université Politechnique de Montréal)

KU Leuven Javier Arroyo

LBNL David Blum

IK4 Tekniker Laura Zabala

3. AGENDA and REPORT

3.1. Approval/ Modification Agenda

Approved

3.2. Status & further discussion where needed

a. Emulators:

4 models selected to start MPC testing

i. BESTEST Air (LBNL/KU Leuven):

documentation to be written and package will be sent to Filip for review

ii. BESTEST Hydronic (KU Leuven /LBNL):

signal exchange blocks have been added, still issues with adding data, to be continued by Filip.

iii. 8-Zone Residential Hydronic (Engie/SDU):

Valentin and Javier collaborated on this, it is BOP-TEST ready, still documentation needed and review (Toa).

iv. Multi-Zone Office Hybrid Simple (KU Leuven/Engie):

Iago finished emulator with RBC, calculation time reduced, meeting to be organized with Valentin.

v. Creating a spot for the emulators:

for now the BOP-TEST repository, maybe later a new repository for the emulators.

b. KPIs - scenarios

representative days/weeks (PNNL), not in the call, postponed.

c. BOP-TEST:

Development needs identified in Rome Meeting

i. Fix issues and/or document solutions (led by LBNL):

1. compilation of test cases that don’t have signal or exchange blocks should receive an error message

2. Docker on Windows generated some issues: documentation needed to get the issues fixed

ii. How to parameterize and implement forecast errors (led by Tekniker):

to be clarified by Dave what is expected – meeting Dave & Jesus

iii. How to parameterize and implement measurement noise (led by Sintef):

postpone to next meeting

iv. How to define workflow for training data for grey-black-box models different than testing data:

For year-long tests need to create “previous” year data, particularly weather data, to train with boundary conditions different than those of the test period. Consider using historic data instead of TMY, or determine how to create new TMY. EnergyPlus has a tool to create TMY data with a subset of measured data, however these should be detailed, difficult to obtain on a broad scale. Dave checks whether Michael has ideas.

v. TAC meeting @ US:

key take-aways will be defined, a lot of interest in BOP-TEST. Funding granted.

d. MPC Library (IbpsaMpc):

i. prototype script and process for Modelica-IBPSA merging (Filip)

3 types of merges were defined

1. direct copy: OK

2. direct copy with parameter changes: OK

3. more advanced changes

whether or not to include examples (as basis for unit tests); IBPSA models have been tested before, they should work regardless of the parameters  merge them without examples (types 1&2)

ii. unit testing framework: first small models (Dave)

specific examples (unit testing) needed for models that have been changed substantially

3. Reporting & Dissemination: start with internal testing (a spreadsheet?)

Internal testing (of whatever controller) on 4 emulators by Canada meeting.

Internal reporting within IBPSA Project 1: what needs to be reported (KPIs, controller characterization)

Database (public): limited information

End user agreement: permission to share the information publicly

First draft of spreadsheet to be made in Canada Meeting based on first controller tests

4. Initiatives joint papers

a. IBPSA Newsletter: contribution of WP1.2

b. MPC Review paper: Dave is reviewing some sections

c. Modelica Conference US: Dave & Michael?

5. Miscellaneous

a. OpenAIGym (https://gym.openai.com/):

Python framework for AI, BOP-TEST could become one of the test environments available on OpenGym.

b. IEA-EBC-Annex 81:

London meeting in September 2019, Dave attended, data for control is one of the topics, focus on development of software to allow testing of controllers is shifted to using data for black/grey model identification in real buildings (system identification based on real data). Discussion will be continued. LBNL might participate.