[TRNSYS-users] controlling the passage to the next time steps (reiter)
pierre.jaboyedoff at effinart.ch
pierre.jaboyedoff at effinart.ch
Mon Jul 12 09:50:36 PDT 2021
Dear Michael,
Thank you so much for your quick answer,
Actually, what i did long time ago in Trnsys 13.1 was the following.
I was simulating a greenhouse with a thermal storage (water tank), the logic
of the control was to initiate each time step with the hypothesis that
nothing is done (neutral, zero control). Then once converged, check if it is
too cold or too warm then going toward controlling the system by iterating
again for each possible case (if too warm, case 1, then 2, then 3, then 4
etc, same when heating is needed. This allowed me to reinitiate the
iteration and convergence until the correct physical case be obtained. Like
that i did have any trouble of convergence to the wrong physical case.
See a short description below. Written in good old f77.
the philosophy is the following:
In my own control routine, at each iteration, I check if the control
configuration is OK. If it is ok then the variable "ireiter" is set to 0,
otherwise it is set to 1, and one start again as if it was the begining of
the same time step.
In my control routine I have the various control cases numbered with the
variable "icas", the variable "icaschk" represents the numbered cases after
check of the physical conditions obtained.
If icaschk does not equal icas after the checks, then I have another
variable called "icorrcas" which is set to either +1 or -1 and will drive
the control to another control case for the next iteration on the same
timestep.
In the routine blockdat.for, I have added two commons. The second /sunsp/ is
actually not needed and was just there to print out variable during
debugging.
-------------------------------------------------------------------------
in blockdat
C MODIF 00000 PJ BEGIN
COMMON /CHKSTEP/ ireiter, icas, icaschk
COMMON /sunsp/ tsunsp, tset, tmax, tload, tsto, text
C MODIF 00000 PJ END
-------------------------------------------------------------------------
Then in main, I have the same commons, I have added a new label 10001 in
order to be able to start the iteration from scratch again, which is checked
with a test on the variable ireiter (if = 0 then go straight, if .gt. 0,
then start to reiter again)
-------------------------------------------------------------------------
-------------------------------------------------------------------------
in main
-------------------------------------------------------------------------
c modif 00000 pj begin
COMMON /CHKSTEP/ IREITER , icas, icaschk
COMMON /sunsp/ tsunsp, tset, tmax, tload, tsto, text
c modif 00000 pj end
-------------------------------------------------------------------------
C ********************** MAIN SIMULATION LOOP **************************
100 IF (DONE) GO TO 999
CALL CLOCK(TIME,DONE,2)
C modif 00000 pj begin
c added label 10001 used for reiter of the time step
10001 IF (NDERIV.EQ.0) THEN
c modif 00000 pj end
-------------------------------------------------------------------------
370 CONTINUE
ENDIF
C SWITCH PREDICTED AND CORRECTED VALUES
I = PRED
PRED = CORR
CORR = I
INTG = 2
IF (IRECUR.GT.0) INTG = 3
IF ((IRECUR.GT.0.OR.NCALLS.GT.0).AND.(ICT.EQ.0)) GO TO 200
C ************************************************************
C MODIF 00000 PJ BEGIN (LABELED COMMON ADDED IN BLOCKDAT_M.FOR, )
c control_iter retuen a variable reiter (in common labelled CHKSTEP)
C REITER =0 GO ON TO NEXT TIME STEP
C =1 REITER ON THE SAME TIME STEP
call control_iter(time)
if (ireiter .gt. 0) then
ireiter=ireiter+1
icasmain=icas
icaschkm=icaschk
timereit=max(0.,time)
c WRITE(*,*) "reiter ", timereit, icasmain, icaschkm
tsunspm=tsunsp
tsetm=tset
tloadm=tload
tmaxm=tmax
stom=tsto
textm=text
c write (*,*) " tsunsp=", tsunspm, " tset=",tsetm," text=",textm
c write (*,*) " tload=", tloadm, " tmax=",tmaxm," tsto=",tsto
c write (*,*) icasmain
c write (*,*) icaschkm
c CALL PRINT(TIME,DONE)
goto 10001
endif
C MODIF 00000 PJ END
-------------------------------------------------------------------------
-------------------------------------------------------------------------
Then in exec.for, one has also modified to set the flag to reiter
-------------------------------------------------------------------------
-------------------------------------------------------------------------
in exec
-------------------------------------------------------------------------
c modif 00000 pj begin
COMMON /CHKSTEP/ IREITER, icas, icaschk
c modif 00000 pj end
-------------------------------------------------------------------------
110 INFO(7,J) = -1
ELSE
DO 111 J = 1, NUNITS
CALL(J) = CALL1(J)
c modif 00000 pj begin
c added info(7,j) =0 for the case where one wants to reiter on the same
c time step
if (ireiter .le. 0) then
INFO(7,J) = -1
else
CALL(J) = .TRUE.
INFO(7,J) = 0
endif
111 continue
c modif 00000 pj end
-------------------------------------------------------------------------
-------------------------------------------------------------------------
In the control routine properly, the variable ireiter is the one that I
check and control.
(TRNSYS type, control routine used)
c modif 00000 pj begin
COMMON /CHKSTEP/ IREITER, icas, icaschk
COMMON /sunsp/ tsunsp, tset, tmax, tload, tsto, text
SAVE /sunsp/
c modif 00000 pj end
it was working very well, and i have now also project where the control
logic sequence may not always follow a simple sequence at each time step.
I had everthing within the TRNSYS code,
I do now know much about ESP-r and its co-simulation, and also does not have
much time for that,
So is there a way to use the type 130 to do the same in TRNSYS 17 ?
If you have any idea or example
Best regards,
Pierre Jaboyedoff,
TRNSYS user since 1987,
Pierre Jaboyedoff
Ing. Thermicien Dipl. EPFL
Effin’Art
Associé
Chemin de Pré-Fleuri 6
Entrée arrière, 3 ème étage gauche
CH-1006 Lausanne
Tel : 00 41 21 616 11 00
Mobile : 00 41 79 668 50 88
pierre.jaboyedoff at effinart.ch <mailto:pierre.jaboyedoff at effinart.ch>
From: Michaël Kummert <michael.kummert at polymtl.ca>
Sent: Monday, 12 July 2021 14:52
To: 'TRNSYS users mailing list at OneBuilding.org'
<trnsys-users at lists.onebuilding.org>
Cc: pierre.jaboyedoff at effinart.ch
Subject: RE: [TRNSYS-users] controlling the passage to the next time steps
Hello,
The co-simulator with ESP-r available in TRNSYS 17 and 18 implements
something that is similar to what you are trying to do, if I understand
correctly.
Once TRNSYS has converged for a time step, Type 130 checks that the
iterations between the 2 programs have converged, and if that is not the
case, Type 130 “forces” TRNSYS to continue iterating at the current time
step.
Perhaps you can take a look at Type 130 and how it interacts with the Kernel
to see whether you can use that for what you are trying to achieve.
The part I am not sure to understand is that in the co-simulation case, it
makes sense to force additional iterations because the inputs coming from
ESP-r will be different so the TRNSYS results after new iterations will be
different. In your case, I am not sure what would cause TRNSYS to find
different results, but I am probably missing something.
In any case, I hope this helps.
Michaël Kummert
Polytechnique Montréal
From: Pierre Jaboyedoff via TRNSYS-users <trnsys-users at lists.onebuilding.org
<mailto:trnsys-users at lists.onebuilding.org> >
Sent: July 9, 2021 08:35
To: trnsys-users at lists.onebuilding.org
<mailto:trnsys-users at lists.onebuilding.org>
Cc: pierre.jaboyedoff at effinart.ch <mailto:pierre.jaboyedoff at effinart.ch>
Subject: [TRNSYS-users] controlling the passage to the next time steps
In trnsys 13.1, i had (long time ago ..) modified the kernel of TRNSYS (see
below).
The reason was to be able to check if the control routine developed was
converging to the correct physical case (e.g. solar, then auxiliary etc
).
If not then i was getting TRNSYS to restart the iteration on the same time
step until the correct control case was obtained.
Has anybody done something similar for TRNSYS 17.2 ?
I have started looking for this, it seems to be located between, exec.for,
trnsys.for, etc
Thanks in advance,
Pierre Jaboyedoff
Head of BEEP PMTU Switzerland (www.beepindia.org <http://www.beepindia.org>
)
Effin’Art
Associate
Chemin de Pré-Fleuri 6
Entrance at the back, 3rd floor on the left
CH-1006 Lausanne
Tel : 00 41 21 616 11 00
Mobile : 00 41 79 668 50 88
pierre.jaboyedoff at effinart.ch <mailto:pierre.jaboyedoff at effinart.ch>
-------------------------------------------------------------------------
-------------------------------------------------------------------------
in blockdat
C MODIF PJ BEGIN
COMMON /CHKSTEP/ ireiter, icas, icaschk
COMMON /sunsp/ tsunsp, tset, tmax, tload, tsto, text
C MODIF PJ END
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------
in main
-------------------------------------------------------------------------
c modif pj begin
COMMON /CHKSTEP/ IREITER , icas, icaschk
COMMON /sunsp/ tsunsp, tset, tmax, tload, tsto, text
c modif pj end
-------------------------------------------------------------------------
C ********************** MAIN SIMULATION LOOP **************************
100 IF (DONE) GO TO 999
CALL CLOCK(TIME,DONE,2)
C modif pj begin
c added label 10001 used for reiter of the time step
10001 IF (NDERIV.EQ.0) THEN
c modif pj end
-------------------------------------------------------------------------
370 CONTINUE
ENDIF
C SWITCH PREDICTED AND CORRECTED VALUES
I = PRED
PRED = CORR
CORR = I
INTG = 2
IF (IRECUR.GT.0) INTG = 3
IF ((IRECUR.GT.0.OR.NCALLS.GT.0).AND.(ICT.EQ.0)) GO TO 200
C ************************************************************
C MODIF PJ BEGIN (LABELED COMMON ADDED IN BLOCKDAT_M.FOR, )
c control_iter retuen a variable reiter (in common labelled CHKSTEP)
C REITER =0 GO ON TO NEXT TIME STEP
C =1 REITER ON THE SAME TIME STEP
call control_iter(time)
if (ireiter .gt. 0) then
ireiter=ireiter+1
icasmain=icas
icaschkm=icaschk
timereit=max(0.,time)
c WRITE(*,*) "reiter ", timereit, icasmain, icaschkm
tsunspm=tsunsp
tsetm=tset
tloadm=tload
tmaxm=tmax
stom=tsto
textm=text
c write (*,*) " tsunsp=", tsunspm, " tset=",tsetm,"
text=",textm
c write (*,*) " tload=", tloadm, " tmax=",tmaxm," tsto=",tsto
c write (*,*) icasmain
c write (*,*) icaschkm
c CALL PRINT(TIME,DONE)
goto 10001
endif
C MODIF PJ END
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------
in exec
-------------------------------------------------------------------------
c modif pj begin
COMMON /CHKSTEP/ IREITER, icas, icaschk
c modif pj end
-------------------------------------------------------------------------
110 INFO(7,J) = -1
ELSE
DO 111 J = 1, NUNITS
CALL(J) = CALL1(J)
c modif pj begin
c added info(7,j) =0 for the case where one wants to reiter on the same
c time step
if (ireiter .le. 0) then
INFO(7,J) = -1
else
CALL(J) = .TRUE.
INFO(7,J) = 0
endif
111 continue
c modif pj end
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------
In my own control routine
c modif pj begin
COMMON /CHKSTEP/ IREITER, icas, icaschk
COMMON /sunsp/ tsunsp, tset, tmax, tload, tsto, text
SAVE /sunsp/
c modif pj end
Pierre Jaboyedoff
Head of BEEP PMTU Switzerland ( <http://www.beepindia.org>
www.beepindia.org)
Effin’Art
Associate
Chemin de Pré-Fleuri 6
Entrance at the back, 3rd floor on the left
CH-1006 Lausanne
Tel : 00 41 21 616 11 00
Mobile : 00 41 79 668 50 88
pierre.jaboyedoff at effinart.ch <mailto:pierre.jaboyedoff at effinart.ch>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/trnsys-users-onebuilding.org/attachments/20210712/4c01980c/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.png
Type: image/png
Size: 11791 bytes
Desc: not available
URL: <http://lists.onebuilding.org/pipermail/trnsys-users-onebuilding.org/attachments/20210712/4c01980c/attachment-0001.png>
More information about the TRNSYS-users
mailing list