# Matched supply additions and trade ranges

2 October 2026. These calculations extend the calibrated September programme. Earlier input editions and results are unchanged.

48 cases use the same 2050 demand, five-year PV calibration, recorded October combined reservoir stock and existing-reactor extensions. Three historical weather years (2011/12, 2016/17 and 2021/22) and two import ceilings (0 and 2 GW) cover seven resource variants. Six further cases replace the 3.2 TWh data-center total with 5 or 10 TWh at the 2 GW ceiling.

| Variant | Changed assumption |
|---|---|
| Reference | Two new 1.1 GW units, Gösgen and Leibstadt, 53 GWp solar |
| Third reactor | Add 1.1 GW, 98% outside a 30-day summer maintenance stop |
| Solar | Add 10 GWp using the same ordinary Swiss hourly profile |
| Battery | Add 40 GWh AC-deliverable energy, 10 GW charge/discharge, 85% round-trip efficiency; empty at both year boundaries |
| Lake | Add 240 GWh pooled capacity; starting water, inflow and turbines unchanged |
| Winter service | Add 1 GW potential output every winter hour; perfect delivery assumed |
| No new reactors | Remove the two new units; add 20 GWp solar and the battery above; retain existing nuclear |

The dispatch first minimizes annual unmet energy, then the largest hourly gap. It retains the earlier import/export preferences only to select a witness. A witness's shortage dates and seasonal split are not uniquely established. Power and energy balances, cyclic storage, turbine limits, boundary stocks and absence of simultaneous charge/discharge or trade are checked after the inherited pumping repair.

The third reactor's maintenance is entirely in summer. Earlier three-unit results used a stop extending into October; those historical results are retained rather than rewritten.

The trade calculations separately minimize and maximize annual exports minus imports while holding annual unmet energy, the peak upper limit and reference annual gross imports fixed. The 8 GW export ceiling remains. These are feasible ranges with unmet demand still present, not forecasts or fully supplied trade plans.

## Main results

- With 2016/17 weather and up to 2 GW imports, the reference still needs 4.643 TWh annually. A third reactor with summer maintenance reduces this to 1.072 TWh. An extra 10 GWp ordinary solar leaves 4.110 TWh. The no-new-reactor solar/battery package leaves 8.167 TWh.
- With 2011/12 weather, the reference peak is 6.148 GW. The no-new-reactor solar/battery package lowers it to 4.069 GW but leaves 6.719 TWh annual unmet electricity against 3.515 TWh in the reference. Energy and power favor different additions.
- Extra reservoir space alone has little value when October water, annual inflow and turbines stay fixed. This does not measure a project's ability to retain more water before October or change inflows and turbine access.
- At the same adequacy and annual gross imports, annual net exports span −8.15 to +11.61 TWh (2011/12), −8.12 to +8.52 TWh (2016/17), and −7.70 to +7.44 TWh (2021/22).
- Data-center totals of 5 and 10 TWh raise the 2016/17-weather annual gap to 5.404 and 7.727 TWh. The flat extra-load profile is an assumption.

The February visual uses only input demand and available generation plus all reservoir/pumped turbine power and permitted imports. Its optimistic ceiling ignores water depletion. In 6–12 February 2012, demand exceeds this ceiling in 50 hours, by 159.43 GWh and up to 6.148 GW. This lower bound does not infer chronology from a chosen seasonal dispatch.

## Limits

These additions differ in size, service and delivery horizon. They establish no least-cost portfolio or cost-per-franc ranking. There is no validated matched wind or alpine-solar portfolio here. Fuel, emissions, owner rights, internal networks, delivery probabilities and site-specific bids are not optimized. The no-new-nuclear package is a serious sensitivity, not an optimized alternative strategy. The 1 GW winter service does not establish a permitted plant or fuel supply.

## Reproduction and validation

From the project root, using its Python environment and preserved earlier input archives:

```sh
.venv/bin/python scripts/review_supply_additions.py --workers 3
.venv/bin/python scripts/review_trade_ranges.py
.venv/bin/python scripts/validate_review_supply.py
.venv/bin/python scripts/build_website_review.py --check
```

Case JSON records include specifications, input totals, solver outcomes, code/input hashes and witness hashes. Hourly witnesses remain in `outputs/review-supply-2026-10-02/`. The validator is read-only. Regeneration needs a new edition rather than overwriting recorded case files. The website evidence edition is `data/website-review-2026-10-02/`.

The public calculation ZIP contains the model code, dependencies and earlier input manifests. The accompanying evidence JSON contains the new case records and displayed calculations. Full historical hourly inputs, repaired hourly schedules and third-party archives remain separate; the ZIP alone cannot reproduce a model run. Their recorded hashes identify the required files.