Repository navigation
Fix/pv handoff harmonic cache - #8
Merged
Merged
Conversation
…tors Carry solved generator reactive power, including Q-limit states and unequal per-phase regulation, into harmonic source initialization, device shunts and current recovery. Preserve gradients and leave caller operating points intact. Distinguish solver-produced reactive-power readout from ordinary prescribed inputs. Symmetric input resolution ignores a user's phase split with a warning; only solved readout preserves the allocation required by nodal balance. Warn when symmetric inputs nevertheless yield unequal solved PV Q, and document the difference between input symmetry and positive-sequence/per-phase regulation. Snapshot prepared-system overrides and branch states by value so replacement, addition, removal and in-place edits cannot silently reuse stale factors. Include cached voltage-base and connection metadata in network validation. Equal-valued replacement tensors retain factor reuse with current gradients. Add PV/PQ equivalence, phase allocation, limit, mutation and differentiation regressions, including native OpenDSS harmonics and pandapower network updates. Co-authored-by: Codex <[email protected]>
Expose explicit harmonic preparation through HarmonicFlowSystem and prepare_harmonic_flow, with simulation and chunked-scenario integration. Recompute operating points, device shunts and injections on every solve. Validate complete assembly dependencies and compare evaluated matrices and numerical options exactly before reusing factors. Rebuild affected entries when parameters, frequency, topology, connection state or defaults change. Bypass harmonic numerical caching for matrix gradients so repeated backward passes use the current graph. Bound retained entries, release obsolete factors before rebuilding, and avoid retaining scenario factors for streaming chunks. Document worker-local ownership, shared-input restrictions and whole-solve locking for callers that need to share a preparation. Add mutation, backend, fusion, solved-Q, gradient, lifecycle and worker-local regressions. Extend native OpenDSS comparisons to the prepared path. Include a standalone synthetic-feeder benchmark requiring no external repository, reference solver, Git executable or dataset; distinguish identical-batch replay from genuinely changed operating points. Co-authored-by: Codex <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.