Brazil 2026 wind/solar curtailment: national exposure and BESS screening boundaries#
News trigger / updated snapshot: BrazilGrid Curtômetro’s Brazil curtailment report, updated 2026-11-07 15:29 Brasília (UTC−03) with ONS-derived observations through 2026-11-06. Underlying primary operator: Operador Nacional do Sistema Elétrico. The included examples/analysis/brazil_curtailment_ons_brazilgrid_20261107.json is a dated, attributed transcription of a secondary publisher’s aggregation, not native ONS raw records, independently recomputed half-hourly values, or measured actual plant generation.
Business question#
How large is the reported opportunity for reducing constraints on Brazil’s wind and PV farms, and what assumptions are required before anyone claims a battery could capture that energy?
Evidence and metric definition#
2026-01-01 to 2026-11-06 (inclusive): BrazilGrid’s restriction-reason-gated, per-group Curtômetro definition reports 23,132.0 GWh wind, 11,718.8 GWh solar, 34,850.8 GWh combined; 326.9 GWh estimated missing-data component is already within the combined total, not additional.
Matched calendar window in 2025: 23,586.9 GWh wind, 10,019.3 GWh solar, 33,606.2 GWh combined, same reporting definition. These are reported potential foregone energy estimates, not metered generated electricity.
November 2026 is only through November 6. October/November and individual months must not be interpreted as complete comparable monthly observation lengths.
BrazilGrid notes that per-plant, per-group and ungated methods produce different totals; this example uses only the per-group restriction-reason-gated+missing-data estimate and never sums incompatible metrics. BrazilGrid’s page is based on ONS open data but a native-source replay has not been performed in this contribution.
Monthly reported figures are shown to one decimal GWh independently of YTD totals, so sums differ by around 0.1 GWh. This is benign display rounding, not an inferred physical measurement error bar or model confidence interval.
Method and outputs#
The reusable energy_methods.curtailment_screen.audit_curtailment checks finite/nonnegative values, unique consecutive months, estimated components as subsets, month-to-YTD reconciliation within the maximum rounding envelope, and year-on-year comparisons on equal cutoff windows. It reports wind share, wind/PV/total YoY changes and the estimated missing-data share. The output chart displays independently reported wind/PV monthly estimates and clearly labels partial November.
A second hypothetical_deliverable_ceiling_gwh screen computes the algebraic upper bound reported curtailed potential GWh × hypothetical fraction chargeable × hypothetical round-trip efficiency. The plot assumes 0–10% chargeable fraction and a hypothetical 90% RTE. It is not a site-level technical maximum and does not demonstrate that any such amount can be captured: it ignores time matching, grid constraints, location, charging power, battery energy capacity, degradation, prices, policy, and whether charging would be permitted. No BESS revenue or investment recommendation follows.
Reproduce and validation#
uv sync --locked --all-packages --group docs --python 3.13
uv run --no-sync pytest methods/tests/test_curtailment_screen.py
uv run --no-sync python examples/analysis/plot_brazil_curtailment_2026.py
uv run --no-sync ruff format --check .
uv run --no-sync ruff check .
uv run --no-sync ty check
uv run --no-sync sphinx-build -W --keep-going -b html docs docs/_build/html
The gallery reads a frozen source summary offline and plots both reported aggregate evidence and separately marked hypothetical scenarios. Tests check arithmetic, rounding, duplicate months, invalid data, inclusion/double-counting and scenario input bounds. A future project-level study should use the repository’s existing ONSClient.wind_curtailment and ONSClient.solar_curtailment, retain native restriction reasons/quality flags and interval timestamps, validate ONS time coverage and calculate feasible site dispatch against actual prices and grid access. This report does not independently confirm BrazilGrid’s aggregation against raw ONS records.