Skip to content

WP1.2 Coordination Meeting September 10, 2021

Jump to bottom Edit New page
Javier edited this page Sep 10, 2021 · 1 revision

Meeting Report WP1.2 ‘Modelica library for MPC’

1. MEETING SUBJECT, DATE

Subject: WP1.2

Date: 10-09-2021

Location: Teams

Minutes taken by: Lieve Helsen (KU Leuven)

2. PARTICIPANTS

3. AGENDA and REPORT

  1. Welcome newcomers
  2. BOPTEST workshop @ BS2021
  3. Joint papers
  4. Forecast uncertainties
  5. Emulators
  6. Expert Meeting – Breakout sessions
  7. Further issues

Welcome newcomers

Laura Maier (RWTH Aachen) and Benjamin Manrique Delgado (SINTEF) join the club, very warm welcome!

BOPTEST workshop @ BS2021

https://bs2021.org/workshops/#1

Core team: Dave (on-line), Javier, Iago, Sen, and Kyle.

Feedback, evaluation and follow-up:

  • 17 onsite attendees + 3 online attendees (most popular workshop that day).

  • It went very well, smoothly from an organizing point of view.

  • Kyle was very instrumental to allow the use of a cloud-based version and to report on the active usage during the workshop.

  • Good engagement, 170 test runs by 20 attendees.

  • Laura Maier enters IBPSA Project 1 thanks to the workshop.

  • Not yet much feedback from the survey sent (Google Form) afterwards.

  • Request for more cases, incl. district level (incl. electrical part). How to enable that expansion? This won’t be part of IBPSA Project 1. Other repositories of emulators to be developed? Quality should be guaranteed, cleaning up and maintenance are needed. Benchmarking remains the main goal. Funded projects are needed to develop the emulators (companies or academics?), typical a limited set of emulators for specific use cases (e.g. HVAC in buildings, energy management of district heating, multi energy vector systems ...). Scalability might become an issue (Francesco Casella is working in this for OpenModelica). Digital twins are being built in multiple projects, should we contact these partners to be involved in our meetings? Call to all: look for funding to develop these emulators.

  • Competition has been announced: challenged to continue using BOPTEST to improve their controller for the hydronic heat pump case, follow-up meeting in 2 months (before the 2nd day expert meeting)

Joint papers

  1. BOPTEST Framework: in revision, very good and positive feedback.

  2. CU and NREL have been wrapping up the MPC and RL paper that leverages the BOPTEST framework (but updated to support the Spawn use case) and they are getting ready to submit it to a journal.

  3. BOPTEST trials paper.

Expand on BOPTEST demonstration and use trials done within this WP, answering questions we would have on MPC (e.g. white/grey/black box approaches, impact of training data, ...)

Forecast uncertainties

Update on the work done by Laura Zabala

  1. Autoregressive model for weather uncertainties (coastal regions ready: Oslo & Berkeley). The next step was to check the behaviour of the error in the temperature forecast in two new inland locations:
  1. A location with inland climate in Norway with the data taken from the previously used Databases. With the help of Harald a couple of locations with the required climate were chosen, data were processed, but not as good as for the case of Oslo, only some isolated days.
  1. In the case of Colorado data, Nicholas couldn’t find the historic predictions, only real historic data are available. Could we have data from somewhere else with inland climate?

  1. Data of Leuven: predictions are updated each 6 hours, this does not agree with the frequency of the other locations, which is of one hour.

  2. Data of Germany: hourly data for 10-days predictions (action Laura Maier).

  3. Data of Milano: historic forecast data available (action Ettore).

  4. Real data as the high uncertainty scenario.

  5. Including the initial error as a user-defined parameter.

  6. Forecast data: how to be combined with observed/measured data? Can these be blended to account for the initial error. Not yet worked out, but on the list of to-do’s. Difference between blended and non-blended could be investigated.

  7. Forecast errors can be as high as 5 degC. Comparative study (KU Leuven) between different forecast services showed that Darksky gives high errors.

Emulators

Updates:

  • Multi-zone (5z) commercial air-based: Dave is addressing the comments, changing the hydraulic configuration of the VAVs, adding nonlinear COP calculations, ...

  • Multi-zone commercial air-based (Yeonjin): Uploaded to Github and waiting for review (action Konstantin)

  • Multi-zone prototype air-based (complex): meeting next week for update (Sen and Iago)

  • Single-zone commercial hydronic: Tao and Konstantin have been working on this, unit test implemented, still some issues; depending on when you start the simulation it ends at some time.

  • Ettore is working on a 2-zone (with different occupancy profiles for the 2 zones) apartment case study in Italy using district heating or heat pump coupled to floor heating, this will be introduced to the BOPTEST repository as well; weather forecast with the uncertainty of Milano could be implemented. Plan:

  • Multi-zone office hybrid (complex): on hold (Filip)

  • Single-zone commercial air-based: 1st version ready by December 2021 (Dave)

  • Multi-zone office hybrid (simple): compilation of envelope OK using PyFMI, with HVAC able to compile but very slow simulation due to large number of events. Trying to reduce the number of events now (Iago & Javier).

Status:

Expert meeting Fall 2021

Dates have been fixed (7-11 Pacific time):

  • Monday October 18

  • Monday November 15

Agenda and link to registration: https://github.com/ibpsa/project1/wiki/Meetings

Breakout sessions (3 or 4) – topics (action Lieve & Dave)

  1. MPC comparison trials for paper + characterizing controllers

  2. With WP3.1 status of DESTEST emulator

  3. Forecast uncertainties: new insights from new data (Germany & Italy) + integrate uncertainties in BOPTEST?

  4. Test results of BOPTEST challenge

  5. Problems with emulators, new KPIs needed,

Further issues

  1. Modelica Conference 2021 (20-24 September)

If you want to learn more, you can attend the - 14th International Modelica Conference. It is an online conference free of charge, but you must register latest September 13. The Modelica Conference covers modeling and simulation of complex physical and cyber-physical systems as well as tools, for a wide range of research and industrial applications.

See: https://2021.international.conference.modelica.org To learn more about Modelica and OpenModelica, you can attend an introductory tutorial Monday Sept 20, 09.00 – 12.30, and an OpenModelica vendor session Tuesday Sept 21, 16.30 – 17.15

Michael Wetter will give a Keynote ‘How can the Modelica community support the transition to decarbonized, grid-flexible buildings?’

Next meeting: October 11, 5 PM-6.30 PM (CET)

Add a custom footer

Wiki

Main project site

https://ibpsa.github.io/project1

Clone this wiki locally