Conversation
pelesh
left a comment
There was a problem hiding this comment.
I think we need to decide if we are going to stick to the modeling approach we defined in the documentation or go back to more compact implementation as in zip load code.
- The model as defined in the README file is larger (has additional 2 algebraic variables) but exposes more parallelism.
- The model as implemented in the C++ code is more compact but requires atomic add to manage system Jacobian assembly.
It would be good to decide what modeling approach to take before merging this PR.
|
Yes, I think the doc approach with Ir and Ii being variables is good. So far I have not been successful in implementing it. I get Enzyme issues. |
What's happening is that in the documented formulation, |
* Alternate Jacobian terms for ZIPLoad. * Jacobian test for LoadZIP. * Fix phasor_load_zip test call. * Minor formatting. * LoadDataZIP --> LoadZIPData in CMakeLists. * Comment out LoadZIP df/dy for now. * Apply pre-commmit fixes * Guard LoadZIPTests.jacobian test with ENABLE_ENZYME. * Expect failure for LoadZIPTests. * LoadZIP initialization. * Fix typo. * Simplify ifrac expression in LoadZIP. * Activate df/dy term in LoadZIP with a hack for null df/dy'. * Apply pre-commmit fixes * Set correct answer key for LoadZIP test * Apply pre-commmit fixes --------- Co-authored-by: nkoukpaizan <nkoukpaizan@users.noreply.github.com> Co-authored-by: pelesh <peless@ornl.gov> Co-authored-by: pelesh <pelesh@users.noreply.github.com>
74e2035 to
44627a4
Compare
|
Thanks all for your help! |
pelesh
left a comment
There was a problem hiding this comment.
All issues fixed, good to merge.
Description
Adding new ZIP load model to phasor dynamics, motivated by the parameter estimation effort.
Proposed changes
Checklist
-Wall -Wpedantic -Wconversion -Wextra.Further comments
None.