search

nikhilgupta10 / GridLAB-D

Other
1 stars 0 forks source link

#662 Convergence Problem in Powerflow module / realtime-server mode, #2402

Closed nikhilgupta10 closed 8 years ago

nikhilgupta10 commented 8 years ago

I am running \GridLAB-D 3.0.0-0 (Hassayampa) 64-bit WINDOWS RELEASE. I am trying to simulate a distribution system that is connected to residential modules in realtime.
I started with a test glm file available in one of the courses named \IEEE_13_house.glm. When I run this file (NOT in realtime) it finishes without errors (the simulation is for one day). When I run the same glm file in realtime, it runs for about 3 hours then it halts with these messages, this happens when I am using FBS (When using NR the simulation does not even start): 

c:\gridlabd-glm\glm-test>gridlabd IEEE_13_house.glm --server
WARNING  [INIT] : gridlabd.conf was not found
WARNING  [INIT] : object name '632-645' does not follow strict naming rules and
may not link correctly during load time
WARNING  [2012-12-12 10:29:16 PST] : last warning message was repeated 28 times
WARNING  [2012-12-12 10:29:16 PST] : deprecate XML usage in request '/zipload1/base_power'
ERROR    [2012-12-12 14:18:52 PST] : convergence iteration limit reached for object node:39
ERROR    [2012-12-12 14:18:52 PST] : last error message was repeated 663 times
ERROR    [2012-12-12 14:18:52 PST] : convergence iteration limit reached at 2012-12-12 14:18:53 PST (exec)
EXCEPTION: convergence failure
ERROR    [2012-12-12 14:18:52 PST] : exec halted: convergence failure
FATAL    [2012-12-12 14:18:52 PST] : environment startup failed: No such file or directory
Model profiler results
======================

Class            Time (s) Time (%) msec/obj
---------------- -------- -------- --------
house              3.348     36.6%      5.3
triplex_meter      2.424     26.5%      3.7
triplex_line       1.504     16.5%      2.4
waterheater        0.890      9.7%      2.0
ZIPload            0.547      6.0%      0.9
collector          0.171      1.9%     42.8
climate            0.123      1.3%    123.0
transformer        0.052      0.6%      2.5
overhead_line      0.040      0.4%      4.4
node               0.026      0.3%      1.5
capacitor          0.005      0.1%      2.5
regulator          0.005      0.1%      5.0
recorder           0.002      0.0%      2.0
underground_line   0.002      0.0%      1.0
switch             0.002      0.0%      2.0
================ ======== ======== ========
Total              9.141    100.0%      3.0

WARNING  [2012-12-12 14:18:52 PST] : last warning message was repeated 37 times
WARNING  [2012-12-12 14:18:52 PST] : new exitcode 8 overwrites existing exitcode 2

I made the following changes to the glm file when I ran it in realtime mode:

  1. removed the start and end time of the clock.
  2. tried removing all schedule_skew statements but the same thing happens.
  3. set minimum_timestep to its default value (which is 1 according to the documentation instead of 60).
  4. set run_realtime to 1.
    I am using the same schedules that are described in the glm files.

,

nikhilgupta10 commented 8 years ago

nikhilgupta10 imported these comments from Sourceforge: "anas83":UPDATE: I tried to simplify the glm file described above (IEEE_13_house.glm). What I did is that I removed all water heaters and removed all temperature schedules and kept only the ziploads at houses. For the ziploads schedules, first, I tried keeping them constant and in another run changing them with time; the GOOD news is that the simulation kept working and I did not get the convergence problem any more.

By doing that (removing water heaters, and removing the temperature schedule), I reduced the load on the system and did not get the convergence problem.

Please let me know if you need the glm files to debug.

Anas

,

"anas83":The previous comment only applies for FBS. NR is still NOT working at all (synchronization problem). Here is what I get when running with NR: 

C:\gridlabd-glm\glm-test>gridlabd IEEE_13_house_new_naming.glm --server
WARNING  [INIT] : gridlabd.conf was not found
WARNING  [INIT] : object name '632-645' does not follow strict naming rules and
may not link correctly during load time
IEEE_13_house_new_naming.glm(593): 0;   // top meter number
ERROR    [2013-01-07 23:11:09 PST] : 2013-01-07 23:11:10 PST: object CAP1 stoppe
d its clock (exec)!
EXCEPTION: synchronization failed
ERROR    [2013-01-07 23:11:09 PST] : last error message was repeated 1 times
ERROR    [2013-01-07 23:11:09 PST] : exec halted: synchronization failed
FATAL    [2013-01-07 23:11:09 PST] : environment startup failed: No such file or
 directory
Model profiler results
======================

Class            Time (s) Time (%) msec/obj
---------------- -------- -------- --------
node               0.284     70.5%     16.7
climate            0.095     23.6%     95.0
collector          0.010      2.5%      2.5
house              0.008      2.0%      0.0
triplex_line       0.003      0.7%      0.0
triplex_meter      0.002      0.5%      0.0
recorder           0.001      0.2%      1.0
================ ======== ======== ========
Total              0.403    100.0%      0.2

WARNING  [2013-01-07 23:11:09 PST] : last warning message was repeated 28 times
WARNING  [2013-01-07 23:11:09 PST] : new exitcode 8 overwrites existing exitcode
 2

,

"jcfuller": * priority changed from undetermined to critical

We ran into the problem of caps stopping their clocks in realtime mode. It was due to an iteration taking longer than the timestep, so the cap was returning a timestamp in the past. I fixed the code in ticket 692, and I've just ported it to 690 (changeset 4110), so it can be merged into main.

Cheers,
Neil

,

"ftuffner": * owner changed from ftuffner to andyfisher

,

"ftuffner": * owner changed from andyfisher to ftuffner

,

"ftuffner": * status changed from accepted to closed

Unable to replicate the issue and user has not responded indicating it is still an issue. Assuming later code revisions fixed this.

,