What
Run and verify a full New York batch using the new subclass RR pipeline: compute_subclass_rr now emits delivery and supply blocks (passthrough / percustomer / epmc / volumetric) and each scenario selects residual_allocation_delivery and residual_allocation_supply. NY should still use percustomer for both on runs 5–6, 9–10, 13–14, so master BAT and master bills for run pairs 1–12 should match a pre-change baseline (within tolerance). Methodology, rationale, and step-by-step commands live in:
context/methods/bat_mc_residual/epmc_and_supply_allocation.md
Why
RI has already been exercised with the new split (e.g. EPMC delivery + passthrough supply on runs 17–18). NY config is unchanged for now, but the shared code path (YAML shape, CAIRO patches, build_master_bat column detection) must be regression-tested on all NY utilities before adopting EPMC on NY delivery or changing HP rate design. The doc also flags expected diffs on demand-flex pairs 13–16.
How
- Confirm Google Sheet has
residual_allocation_delivery / residual_allocation_supply; NY subclass runs use percustomer for both where applicable.
just s ny all-pre and confirm ny/config/rev_requirement/*.yaml has nested delivery: / supply: with all methods; percustomer delivery should match prior scalar delivery values.- Run NY batch for all utilities (e.g.
run-all-parallel-tracks with agreed RDP_BATCH name).
just s ny build-all-master <batch>.- Execute parity checks from doc §3 (“LLM verification” block): new columns (
BAT_epmc_*, no BAT_peak_delivery where removed), master BAT/bills join vs baseline batch for pairs 1–12 within 1e-6, document 13–16 diffs as expected if present, building counts consistent.
Deliverables
- One completed NY batch name (e.g.
ny_YYYYMMDD_r1-16_*) recorded on the issue.
- Short note or comment: checks 1–3 and 5 pass for pairs 1–12 vs chosen baseline batch; check 4 outcome summarized for 13–16.
- Any anomaly filed or linked as a follow-up (separate issue if scope grows).
cReferences
context/methods/bat_mc_residual/epmc_and_supply_allocation.md (EPMC definition, supply residual open questions, supply EPMC subtraction bug, RI vs NY run tables).- Related future work (optional separate tickets): fix supply-side EPMC in CAIRO (decompose EB per customer); NY policy choice on delivery EPMC and supply allocation.
What
Run and verify a full New York batch using the new subclass RR pipeline:
compute_subclass_rrnow emits delivery and supply blocks (passthrough / percustomer / epmc / volumetric) and each scenario selectsresidual_allocation_deliveryandresidual_allocation_supply. NY should still use percustomer for both on runs 5–6, 9–10, 13–14, so master BAT and master bills for run pairs 1–12 should match a pre-change baseline (within tolerance). Methodology, rationale, and step-by-step commands live in:context/methods/bat_mc_residual/epmc_and_supply_allocation.mdWhy
RI has already been exercised with the new split (e.g. EPMC delivery + passthrough supply on runs 17–18). NY config is unchanged for now, but the shared code path (YAML shape, CAIRO patches,
build_master_batcolumn detection) must be regression-tested on all NY utilities before adopting EPMC on NY delivery or changing HP rate design. The doc also flags expected diffs on demand-flex pairs 13–16.How
residual_allocation_delivery/residual_allocation_supply; NY subclass runs usepercustomerfor both where applicable.just s ny all-preand confirmny/config/rev_requirement/*.yamlhas nesteddelivery:/supply:with all methods;percustomerdelivery should match prior scalar delivery values.run-all-parallel-trackswith agreedRDP_BATCHname).just s ny build-all-master <batch>.BAT_epmc_*, noBAT_peak_deliverywhere removed), master BAT/bills join vs baseline batch for pairs 1–12 within 1e-6, document 13–16 diffs as expected if present, building counts consistent.Deliverables
ny_YYYYMMDD_r1-16_*) recorded on the issue.cReferences
context/methods/bat_mc_residual/epmc_and_supply_allocation.md(EPMC definition, supply residual open questions, supply EPMC subtraction bug, RI vs NY run tables).