-
Notifications
You must be signed in to change notification settings - Fork 14
2017 07 tag agenda
Participation at this meeting is by invitation only.
Date: July 19, 2017, 11am-1pm PDT
Slides are posted at https://github.com/lbl-srg/obc/tree/master/meetings/2017-07-tag
See below agenda for dial-in information.
- Status update: Upcoming deadlines, see https://github.com/lbl-srg/obc/wiki/2017-07-tag-next-steps
- CDL: Need review and control vendor input.
- G36 implementation: Need review.
- Case study. Need a team member to identify building and contribute to study. For this case study, need to work on secondary system only. A small system in terms of number of control loops, and a small building would be best suited. The system should be well documented. We like to share results from the case study in the open literature (the building could be made anonymous if needed). (Primary system and facade will come later.)
- Amir Roth
- Brent Eubanks
- Gerry Hamilton
- Charles Holleran
- Jianjun Hu
- Jon Schoenfeld
- Karen Perrin
- Marina Sofos
- Milica Grahovac
- Paul Ehrlich
- Paul Switenski
- Steven Taylor
- Philip Haves
- Jay Santos
- Janie Page
- Michael Wetter
LBNL team obtained feedback from Larry Bacher, ALC, to date. Mark, Paul, Brent, Jon, and Jay offered to reach out to vendors for further feedback. The feedback is welcome in plain text and word format. You could also click https://github.com/lbl-srg/modelica-buildings/issues/new to report any issues in which case the discussion will be tracked for all to see. The goal is to be able to support all known sequences by different vendors, either natively or via composition of basic blocks, although the project will only have resources to implement some sequences.
The team went back to this issue later in the meeting and discussed options for cases where the vendor control coding language has a) extra blocks, b) lacks our blocks. The team also suggested to discuss the above and the generation of C code for the execution on systems that allow use of (compiled) C code. The concern is that line programming vs block code may be significantly different in vendor languages. Phil noted philosophical distinction PPCL (Siemens) vs function block graphical language. Declarative vs imperative camps of coding noted. Michael noted that imperative programs can be generated from declarative, which is what Modelica tools are doing, but the other way around is not possible. Workshop in August got set as a target for the meeting.
Note: Further pointers are at http://w3.usa.siemens.com/buildingtechnologies/us/en/education-services/building-automation/building-automation-self-study/pages/introduction-to-control-programming.aspx and https://www.quia.com/files/quia/users/hpiracer/AIRC65/PPCL-Intro_MTSAC.pdf
No comments were received on the topic of basic blocks composed of other basic blocks.
CDL to be exposed as needed to users, not whole code base.
We discussed how to address the situation that some sequence are virtually identical but may take different inputs and parameters if there is heating and/or humidification and/or cooling. Options are to dublicate and modify sequences in the library, or use of a boolean flag to conditionally remove parts that are not needed, or connecting dummy inputs to non-needed inputs. Brent suggested a list of questions presented to the user, which essentially define the value of the boolean parameters that are used to turn certain features of the block on or off. He strongly suggested to avoid adding separate sequences to avoid explosion of the size of the library. Paul agreed and Michael agreed. Brent suggested to extend that functionality to include possibility of adding elements, such as humidifier, which are currently not a part of G36. This is related to the scoping issues discussed in the following section.
CLD does not implement it. The team agreed that it is not necessary.
Collect feedback from control vendors on the scope of the CDL basic block library by September 15.
LBNL team intends to release first version of CDL by September 30 (end of Q4).
Optimal start-up - most vendors keep their algorithm proprietary. Discussion on how other control vendors retrieve the operation mode signal: Information provided by several external project partners. If building switches from occupied to unoccupied to occupied again, when/how does controller know about occupation again? Vendors use different approaches: ALC and G36 starts at zone level which then informs system. Question of how much ahead the warm-up needs to start. Many systems have schedule control points at which operation changes can be implemented. Can also have some influence by recent historical operations (e.g. recursive on outside air temperature). Most systems have learning based implementations. Operationally, optimal start differs from occupancy. Occupancy, for example, will include ventilation. Need hooks in CDL code to attach to vendor specific optimal start blocks. Units should have flags to indicate assumed state (e.g. optimal start, occupied, unoccupied). LBNL team to implement an approximation to optimal start up that will allow control sequence testing and verification.
The team agreed to using G36 PPR1 as our reference version of the guideline, being cognizant of elements expected to change. Brent emphasized that some of the sequences, e.g. minimal ventilation logic, might undergo significant changes in future instances of G36.
LBNL team demonstrated multi-zone VAV AHU economizer outdoor and return air damper limit sequence from G36 with plotted results.
Tags in CDL – trying not to pick any particular scheme, but be aware of Brick and Haystack as relatively large players in the area.
LBNL team requested validation and review on the level of atomic and composite sequences. The team discussed the best approach. Options for testing:
- Inspect code and building blocks (parse at visual logic level)
- Use test harness to confirm outputs for given input (the models run with OpenModelica's graphical editor OMEdit, https://openmodelica.org, or Dymola or JModelica (from a Python shell, but no GUI))
- Connect sequences to model of air handler and simulate under different seasons – works without Dymola using open source tools
The team agreed to organize a meeting with developers and key partners to obtain guidance on the scope of the G36 CDL library. Brent suggested Integral offices for the meeting and Jianjun offered to make a doodle for the meeting.
Deadlines: Mid September - release a first version of the library
Structure (atomic, composite) useful to sequence implementation? Better form structure for sequences library?
Jianjun is looking at alternate structure that may allow easier reuse.
In August. The team agreed to identify suitable experts, ask for feedback, and organize a workshop. Main questions are whether blocks and language can be converted to different vendor systems, or supported by vendors if not already present, and what the critical corner-cases are.
In August. The team agreed to organize a workshop at Integral Group offices.
The team needs to identify a suitable real building for a case study, ideally multizone. Potential sources: Steve mentioned RP 1747 (E+ and Contam). Phil offered to contact Joe Zhou.
Ideally looking for building that can be used as a case study in simulation mode, not actual closed loop on real building. Look at HVAC performance when controlled using project code: Single or multiple zone AHU in Modelica using G36, then add building model to emulate load, compare differences in energy use. Use building with a few zones, ideally 2 to 5, but not hundreds. Ideally LBNL designers work with other team members to show how to use sequences. Gerry Hamilton notes that Stanford's simplest building has 50 zones, but may have some simple components. Could be AHU with few zones. Looking for VAV with reheat. Steve Taylor not sure he has building in his portfolio that works. Goal is prove usability of CDL for actual sequences, although for now, study will be in simulation only as CDL export and translation tools have not yet been developed. May not need actual building at this stage. Noted a VAV system with about 4-5 AHUs that may be available for consideration. Or use FLEXLAB? Other options identified as well. Simulation study at Iowa Energy Center proposed since well documented, has trend data. RP1747 may also provide insights. Don’t want to have to implement special control sequences just for the test.
Join from PC, Mac, Linux, iOS or Android: https://lbnl.zoom.us/j/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