2026 State RPS / CES /Hydrofrac Update

[Yunzhi-Chen]

Summary

Annual update of state RPS / CES / voluntary policy inputs in state_policies/ to the June 2026 LBNL release of "RPS & CES Targets and Demand". The three CSVs ReEDS consumes — rps_fraction.csv, ces_fraction.csv, and hydrofrac_policy.csv — are regenerated from the new data.

Technical details

Two upstream data sources were refreshed:

1. LBNL state RPS / CES dataset (Galen Barbose):
   RPS & CES Targets and Demand_June 2025.xlsx → RPS & CES Targets and Demand_June 2026.xlsx.
   The June 2026 release renamed several sheets and shifted header rows:
   • Statewide Load → Statewide Sales
   • RPS & CES Demand Projections → RPS & CES Demand (GWh)
   • LBNL also removed the Non-RE Accounting sheet from the public workbook. Galen sent it directly to Wesley, and it was pasted back into the workbook as a sheet named Non-RE Accounting so the script's input is self-contained.
2. NLR Green Power Data (used for the voluntary RPS row):
   nrel-green-power-data-v2023.xlsx → nrel-green-power-data-v2024.xlsx. The 2024 release added one more historical year (2024) and revised earlier years' market-wide estimates downward, which propagates into a uniformly lower voluntary trajectory in rps_fraction.csv.

Implementation notes

Changes in state_policies/data_processing.py are all configuration at the top of the file — no algorithmic changes:

• filename → RPS & CES Targets and Demand_June 2026.xlsx
• filename_voluntary → nrel-green-power-data-v2024.xlsx.
• Voluntary historical-year cap in the projection block bumped from <= 2023 to <= 2024 so the new 2024 data point is included in rps_series rather than projected; projection still extrapolates to 2050 using the minimum year-over-year growth in the historical series.
• base_year bumped from 2023 to 2025. Years strictly before base_year are passed through as raw values from the LBNL workbook, while years ≥ base_year go through the change-point detection + linear interpolation + monotonic max pipeline.
• hydro_year (for gen_ann.csv) kept at 2023 to stay time-consistent with the vintage of Galen's Non-RE Accounting numerator data (2022–2023 reports).
• gen_ann.csv and hierarchy.csv are updated using recent ReEDS run from ReEDS main at tag 2026.04.17, commit 62f6381ec3da9005f0d7520dd37db235fc4c8784.

Additional changes

• Rewrote state_policies/README.md.
• Replaced old and new data comparison/old and new data comparison.ipynb with old and new data comparison/generate_comparison_plots.py, which produces standardized PNGs from the old and new outputs.
• Staged the previous run's outputs in old and new data comparison/old ReEDS input/ so the diffs/plots are reproducible by reviewers.

Issues resolved

Closes #122.

Relevant sources or documentation

• LBNL Renewables Portfolio Standards project (annual dataset and accompanying summary slides): https://emp.lbl.gov/projects/renewables-portfolio
• NLR Green Power Data (voluntary procurement, formerly NREL): https://www.nlr.gov/analysis/voluntary-power-procurement
• The LBNL June 2026 summary slides explain the RPS changes for AZ, CT, ME, NC, NY and VA.

2025 vs 2026 comparisons

Additional details

• Output verification: Ran python data_processing.py to regenerate all six before/after PNGs in old and new data comparison/plots/. Spot-checked notable state-level changes against the LBNL summary slides (e.g., the CT RPS decline traces to the Class I / LFG / biomass eligibility change cited above). Voluntary trajectory verified; the uniform downward shift matches NLR's revised market-wide estimates in v2024.
• LLM usage: Used GitHub Copilot (chat) to assist with editing data_processing.py for the new LBNL sheet layouts, drafting the README's annual-update procedure, writing generate_comparison_plots.py, extracting text from the LBNL summary-slides PDF when investigating the CT decline, and locating the v2024 NLR file. All resulting code/text was reviewed and the regenerated CSVs were spot-checked against the source workbooks before committing.

[wesleyjcole]

The IL CES fraction ramp looks to be off in 2040-2050. Would you look into that to confirm if it is actually correct or if it needs adjusted?