CED saved-prediction replay for Supplement A24.
Run: python replay.py
Python standard library only. Reads sealed forecast/target CSVs and parameters; verifies all metrics and fixed-history identity. No training, model execution, new forecasts, physical integration or independent-data generation.
Both evaluation records were previously examined. Future recorded controls are available; the mirror is not byte-verified against the author archive. Logical read counts are not measured energy savings.
Original dataset/benchmark: Wigren and Schoukens (2017), https://www.nonlinearbenchmark.org/benchmarks/coupled-electric-drives . Original report and raw MAT payloads are not redistributed in this companion. Saved evaluation arrays and derived forecasts do not reconstruct the full model-development or hardware experiment.
This is a local revision companion, not a new public deposit.
