Skip to content

Add justmip/SoCDRv2 fixes to main - #701

Open
alextsimp wants to merge 7 commits into
mainfrom
justmip-for-main
Open

alextsimp wants to merge 7 commits into
mainfrom
justmip-for-main

Conversation

@alextsimp

Copy link
Copy Markdown
Member

Fix 1 — Car energy doubles after 2050 despite a 90% BEV stock

Three causes compound:

  1. Energy per km is frozen at the base year. For PC, V01ConsSpecificFuel is fixed (.FX) from i01SFCPC(…,"%fBaseY%") (preloop.gms#L49), and i01SFCPC is copied forward unchanged (input.gms#L144). No powertrain has any autonomous efficiency improvement.
  2. The region-wide calibration scaler distorts BEVs. The base-year scaler is fitted to total PC fuel and multiplies every technology, BEVs included. As a result, MEA's SFC is 4–5× that of other regions (GSL 526, ELC 92.5 vs ≈ 14 in DEU), which means a BEV uses ≈ 1 kWh/km. Likely causes: MEA's base-year PC fuel includes non-car road use, or its stock data is too low.
  3. Saturation is a uniform 700 cars/1000 people, and km/veh grows with income. i01PassCarsMarkSat = 0.7 for every region (input.gms#L224). The km/veh equation has a hard-coded GDP/cap elasticity of 0.2, and its fuel-price terms are commented out (equations.gms#L229-L232). So km/veh rises from 12.3k to 19.8k, and carbon prices never affect car activity. The world fleet grows from 1.2 to 4.65 bn.

All growth is in developing regions. Car FE, 2020 → 2100: IND 68 → 721, MEA 146 → 695, SSA 50 → 594, OAS 141 → 559. Every OECD region falls. That regional fleet growth is expected. Combined with frozen, oversized per-km consumption, though, it gives ≈ 28,000 TWh of BEV electricity in 2100.

Added all fixes but as comments. The CurPol emissions for cars are reduced too much if applied. More investigation is needed for transport. Use the fix if you want to reduce the CO2 transport emissions.

Fix 2 — Data-centre demand grows exponentially in the input, even with ICT = Lower

The world sum of data/iFuelConsICT.csv (SSP2) goes from 117 Mtoe (2050) to 307 Mtoe (2070) to 1,297 Mtoe (2100), which is ≈ 15,000 TWh. That is ≈ 5%/yr growth, doubling every ~14 years, with no saturation. Central/Mean reaches ≈ 4,200 Mtoe and Upper 7,768 Mtoe. This is an extrapolation artefact of the econometric fit in mrprom, not a model response.

Change the year if you want to reduce the very large growth of ICT. Now it is deactivated.

Fix 3 — Energy and waste CH₄ are not linked to the model's fossil activity

Q07EmiActBySrcRegTim = i07DataCh4N2OFEmis (SSP2 baseline, Harmsen et al. 2019, Mt C-eq) − V07RedAbsBySrcRegTim. The second term is the highest MAC step whose cost ≤ the carbon price (modules/07_Emissions/legacy/equations.gms). Nothing in the equation reads coal, oil or gas production, and the exogenous baselines move the opposite way to the model:

2100 vs 2020 Model primary energy Exogenous CH₄ baseline
Coal ×0.20 ×2.37 (249 → 591 Mt C-eq)
Gas ×0.71 ×1.64
Oil ×0.41 ×0.80

So the model phases out coal, yet coal-mine CH₄ is abated from a baseline that more than doubles. Scaling each fugitive baseline by the model's own fuel activity (ratio to its 2020 level, per region) would cut 2100 energy CH₄ from ≈ 77 to ≈ 28 Mt (−49 Mt). This is the main fixable gap.

Fix 4 — ✅ Applied (option A): solar thermal and geothermal were nearly free

TSOL and TGEO copied the heat-pump cost row, paid no fuel, had a variable cost of 1e-3 and used utilisation 1. Their useful-heat cost was 0.09–0.24, against 0.4–0.6 for heat pumps and 1–4 for gas and electricity. The building calibration had already pushed their maturity to the 0.01 floor and still could not hold them back. In AG, their maturity is 1.

Changes

  • Q02CapCostTech (and its preloop copy for historical years) now divides capital + fixed cost by i02util as well as by conversion efficiency. i02util = 1 for every other technology, so nothing else changes. Equipment and fuel accounting already used i02util.
  • i02util: TSOL 0.12, TGEO 0.40 in DOMSE (02_Industry/technology/input.gms).
  • TSOL and TGEO stay available in agriculture (greenhouses, drying).
  • CHA 2030 useful cost: TSOL 0.09 → 0.76 (SE/AG) and 0.12 → 0.98 (HOU); TGEO 0.18 → 0.83.
  • Heat pumps (0.41 SE / 0.32 HOU) are now the cheapest heat option.

World FE, Mtoe, HOU + SE + AG (without → with):

2030 2040
Solar thermal 404 → 42 834 → 43
Geothermal 263 → 31 456 → 49
Electricity 1,042 → 1,114 1,401 → 1,481
Natural gas 569 → 737 352 → 771

⚠️ Gas absorbs much of what solar loses, especially in AG (5 → 52 Mtoe by 2040). There, TNGS costs only 0.14–0.19 and every technology has maturity 1.

Results -16/11/2026

CH4 CO2 CO2_Demand_Supply Kyoto ## With Fix 1 for car transports CO2_all_transport

@alextsimp alextsimp self-assigned this Nov 16, 2026
@alextsimp alextsimp linked an issue Nov 16, 2026 that may be closed by this pull request
5 tasks done

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

JustMIP scenarios

1 participant