ESDC1A Documentation + Implementation

[lukelowry]

Description

Implementation and documentation for the ESDC1A PhasorDynamics exciter model.

Proposed changes

• Added ESDC1A model implementation, documentation, and diagram.
• Added ESDC1A JSON parsing and SystemModel assembly support.
• Added focused PhasorDynamicsExciterEsdc1aTest unit coverage.
• Added sparse Jacobian support through Enzyme and DependencyTracking build paths.
• Updated exciter registration, input-format docs, and changelog.

Checklist

• All tests pass (except known PE failure)
• Code compiles cleanly with flags -Wall -Wpedantic -Wconversion -Wextra.
• The new code follows GridKit™ style guidelines.
• There are unit tests for the new code.
• The new code is documented.
• The feature branch is rebased with respect to the target branch.
• I have updated CHANGELOG.md to reflect the changes in this PR.

Further comments (A)

@nkoukpaizan, I keep running into an issue with Enzyme for Exciter models in the control voltage effective input algebraic equation:

const ScalarT ref_err = vref_ - ev;
const ScalarT sig_err = vs - vc;
const ScalarT lim_err = sUEL_ * vuel - vf;
f[EV]                 = ref_err + sig_err + lim_err;

Enzyme fails if that is an inline expression. Sparsity issue. I am not sure if this is intentional, but I finally found a reliable way to circumvent the issue within the model implementation (the code above).

Further comments (B)

@pelesh @abirchfield @alexander-novo

Out of PR scope, but there is a modeling issue that I'd like to raise here:

• The 'Compensation voltage', $E_C$, typically comes from the machine model instead of the terminal Bus. This is because there is a parameter specified by the machine model (e.g.GENROU) that we do not currently support: $R_{comp}$ and $X_{comp}$. When these are both zero-valued, $E_C=|V_{bus}|$. Sometimes, the compensation reactance is configured to be nonzero specifically to decouple $E_C$ from $|V_{bus}|$ and improve numerical stability. I think there is a broader conversation to be had about how we can model sensed terminal voltage.
• This exciter as well as IEEET1, locally compute $E_C=\sqrt{V_r^2+V_i^2}$, despite $E_C$ being documented as an external variable. This was a temporary design choice before we had signals.
• For more advanced grid cases and models, the terminal voltage magnitude is consumed more than once. Therefore, I encourage us to introduce $V_{m}$ as a third internal algebraic variable of Bus, conditionally included based on consumption (otherwise we could introduce unused nonlinear algebraic equations incessantly!). Same for $\theta_{bus}$. I will soon need models that need the bus frequency (i.e., time derivative of $\theta_{bus}$).

Please let me know if I should open distinct issues for these. They are front of mind as I implemented this so I am leaving here for now.

[lukelowry]

Rebased and cleaned. Issue discussed in PR body is still an issue but out of the scope of this PR

[lukelowry]

The model is ready @abirchfield @nkoukpaizan

[abirchfield]

Reviewed and discussed in person. Ready to merge.

[lukelowry]

Rebased!

[nkoukpaizan]

Good enough to merge. Thanks for the hard work!

[lukelowry]

Good to go. I'll begin making the other open PR models align with this and REGCA. Not opposed if you want to merge #521 after this

[pelesh]

My prior request is addressed.