Switch Object with NR solver

[MA]

Hello,

I hope all is well.

Reference : The file attached is obtained and created by Cody from a post here in sourceforge.

GridLAB-D Version:

GridLAB-D 4.3.0-18551 (Navajo [c91d5848:develop:Mod]) 64-bit LINUX RELEASE

I'm working on a project where I control waterheater by switch. The waterheater is driven by water demand schedule. When the switch is open, I'm assuming the water temperature will reduce but the waterheater won't be able to turn on to heat the water to the setpoint specified in the waterheater object.

Since switch object may not behave properly under FBS, I used NR in IEEE 4 node feeder. I used switch object on the triplex portion of the feeder, between two triplex meters (from meter_A_1 to Load_Meter). Also, I used fault_check object.

I should mention that the file works just fine with FBS with switch object included. It also works well with NR solver without switch object. That being said, when I run the glm file in NR with switch object included, I get the following error:

ERROR    [2018-06-20 00:00:00 EST] : From node has no voltage.
ERROR    [2018-06-20 00:00:00 EST] : commit_all() failure: object impedance_dump:2 commit failed
ERROR    [2018-06-20 00:00:00 EST] : model_commit_failed
ERROR    [2018-06-20 00:00:00 EST] : exec halted: commit failure
FATAL    [2018-06-20 00:00:00 EST] : shutdown after simulation stopped prematurely
FATAL    [2018-06-20 00:00:00 EST] : environment startup failed: Success
WARNING  [2018-06-20 00:00:00 EST] : last warning message was repeated 1 times

I ran the file in verbose mode, and here's the result:

   ... load time: 0 sec
   ... starting up batch environment
   ... initializing objects...
   ... installation type not specified
   ... External solver solver_KLU found, utilizing for NR
WARNING  [INIT] : house_e: no climate data found, using static data
WARNING  [INIT] : waterheater::init() : height and diameter were not specified, defaulting to 3.78 ft
   ... waterheater::init() : setting initial height to tank height.
   ... wh1 (waterheater:32) schedule is not specified so the load may be inactive
WARNING  [INIT] : house_e: no climate data found, using static data
WARNING  [INIT] : house_e: no climate data found, using static data
   ... creating index 0
   ... creating index 1
   ... creating index 2
   ... shuffled 7 lists in index 0
   ... shuffled 7 lists in index 1
   ... shuffled 7 lists in index 2
   ... switch_object::sync:switch1:1529470800:1:1:1
   ...   switch_sync_function:switch1:1:1:1:1:1
   ...   BOTH_switch_sync_pre leaving::
   ...   calling link_object::sync:1:1:1
   ...   NR_switch_sync_post::
   ... External LU solver failed out with return value -3
ERROR    [2018-06-20 00:00:00 EST] : From node has no voltage.
ERROR    [2018-06-20 00:00:00 EST] : commit_all() failure: object impedance_dump:2 commit failed
ERROR    [2018-06-20 00:00:00 EST] : model_commit_failed
ERROR    [2018-06-20 00:00:00 EST] : exec halted: commit failure
FATAL    [2018-06-20 00:00:00 EST] : shutdown after simulation stopped prematurely
FATAL    [2018-06-20 00:00:00 EST] : environment startup failed: Success
   ... shutdown complete
   ... elapsed runtime 0 seconds
   ... exit code 2

My questions are:

1- Which node is this error referring to?

2- Are there specific steps to be taken when using switch object with NR solver? or am I missing something else?

Any suggestion would help.

Thank you

[MA]

UPDATE:

References:
http://www.pnl.gov/main/publications/external/technical_reports/PNNL-17694.pdf

http://gridlabd.sourceforge.net/wiki/index.php/Market_User_Guide#Passive_Controller_Object

Following the second link, I used passive controller to turn ON/OFF the water heater. It's behaving as expected.

Regarding the first post, I still can't figure out why the water heater wouldn't behave well with switch object.

Any help would be appreciated.

[Frank Tuffner]

Hello MA,

I'm a bit concerned when you say it works fine in FBS with the switch, but not NR. It is failing for the same reason in both solvers for me, which it should. The error is associated with the impedance_dump object not being able to map a property, due to the switch being on a triplex/split-phase system. Switches do not support the triplex/split-phase implementation at this time - they only work on the three-phase portion of the system (i.e., they can be single-phase, but not split-phase, so no "S" in the phases property).

In terms of the overall behavior you are trying to model, I highly recommend using things like the passive_controller or something in the market module. Switching the power directly via the powerflow isn't a typical use case scenario, so I'm not sure the water heater will behave properly. If it is switched with the passive_controller (or maybe the object's breaker), that has been tested and should work.

-Frank

[MA]

Hi Frank,

Im a bit concerned when you say it works fine in FBS with the switch, but not NR. It is failing for the same reason in both solvers for me, which it should.

I went over the glm file with FBS solver. It didn't have impedance_dump object in it. The glm file with FBS solver ran without any errors but the water heater didn't respond at all to the switch state. After reading your response, I understand the reason behind the waterheater behavior.

In terms of the overall behavior you are trying to model, I highly recommend using things like the passive_controller or something in the market module. Switching the power directly via the powerflow isn't a typical use case scenario, so I'm not sure the water heater will behave properly. If it is switched with the passive_controller (or maybe the object's breaker), that has been tested and should work.

I did use passive_controller object and the water heater reponded accordingly.

Thank you for your response!