Sector-coupled model (draft)部門耦合模型（草稿）

Why the first runs are not Taiwan's pathway為何最初的模擬不是台灣的路徑

The first runs (the early test and sensitivities A–D) allowed unlimited new nuclear and could not import hydrogen or other low-carbon fuels, so they built large nuclear fleets. That is not a realistic picture of Taiwan: its 2050 net-zero pathway has no new nuclear and relies on imported fuels. Those runs only tested the model. PyPSA-Earth's sector-coupled model also lacks options the official plan relies on; this fork adds them and runs the official pathway (below).最初的模擬（早期測試與敏感度 A–D）允許不設上限的新核電，且無法進口氫氣或其他低碳燃料，因此建置了大量核電。這並不符合台灣的實際情況：台灣 2050 淨零路徑沒有新核電，並仰賴進口燃料。這些模擬僅用於測試模型。PyPSA-Earth 部門耦合模型也缺少官方規劃所依賴的選項；本分支加入這些選項，並模擬官方路徑（見下方）。

Official figures checked in the approved key-strategy action plans (April 2023), which quote the NDC pathway; the 2050 CO₂ storage volume is not in the CCUS plan's text and is still to verify.官方數字已與核定之各關鍵戰略行動計畫（2023 年 4 月，引述國發會淨零路徑）核對；2050 年 CO₂ 封存量未見於碳捕捉利用及封存行動計畫內文，仍待查證。

Gaps found in PyPSA-Earth's sector-coupled model for TaiwanPyPSA-Earth 部門耦合模型用於台灣時發現的缺口

1. No gas power plant with carbon capture (only district-heating CHP, which Taiwan hardly has).沒有燃氣發電搭配碳捕捉的電廠（僅有區域供熱汽電共生，而台灣幾乎沒有區域供熱）。
2. No hydrogen or ammonia imports (export only); added in this fork for sensitivity C.沒有氫氣或氨的進口（僅有出口）；本分支為敏感度 C 新增此選項。
3. Missing Taiwan data that silently distorts results: no demand growth rates (the default multiplies household demand about fivefold by 2050), zero cars (no EV chargers), and Taiwan listed as "Other Asia" in the UN energy data.缺少台灣資料，會默默扭曲結果：沒有需求成長率（預設值使 2050 年住宅用電增為約五倍）、汽車數為零（沒有電動車充電）、聯合國能源資料將台灣列為「Other Asia」。
4. Bugs in the multi-year (myopic) chain, fixed in this fork: existing plants never retired, land-use limits counted existing capacity twice, converted biomass plants got unlimited free fuel, and clashing names dropped all links.多年期（短視）流程的錯誤，已在本分支修正：既有電廠從不除役、土地利用上限重複扣除既有容量、轉換後的生質能電廠獲得無限免費燃料、名稱衝突導致所有連結遺失。
5. European defaults: 200 Mt/yr CO₂ storage, no floating offshore wind, zero geothermal investment cost.歐洲預設值：每年 200 Mt CO₂ 封存、沒有浮動式離岸風電、地熱投資成本為零。

First step: Taiwan's official pathway in this model第一步：在本模型中建立台灣官方路徑

The fork now offers what the official plan relies on: gas combined cycle with carbon capture (NETL cost and efficiency), hydrogen and ammonia turbines, and imports of hydrogen, ammonia, synthetic oil and synthetic methane, priced after Hampp et al. (2023). Offshore wind (40–55 GW), solar (40–80 GW) and geothermal (3–6.2 GW) are held to the action plans' 2050 ranges, and new nuclear is capped at zero unless stated. Pick the runs in the scenario menu below.本分支已加入官方規劃所依賴的選項：燃氣複循環搭配碳捕捉（採 NETL 成本與效率）、氫能與氨渦輪機，以及氫氣、氨、合成燃油與合成甲烷的進口（價格依 Hampp 等人 2023 年研究）。離岸風電（40–55 GW）、太陽光電（40–80 GW）與地熱（3–6.2 GW）限於行動計畫的 2050 年範圍；除另有說明外，新核電上限為零。可於下方情境選單中選取各模擬。

• With the official options and imports, net zero in 2050 is reached without new nuclear and without unmet demand.採用官方選項與進口後，2050 年不新建核電也能達成淨零，且沒有未供應的需求。
• When every carrier can be imported, the model buys about 310 TWh a year: synthetic methane 120, ammonia 93 (all of it burned in ammonia turbines for power), synthetic oil 92 and almost no hydrogen. The earlier result of 300–440 TWh of imported hydrogen came from hydrogen being the only carrier the model could import; the amount of imported energy is what holds.當所有能源載體都可進口時，模型每年進口約 310 TWh：合成甲烷 120、氨 93（全部用於氨渦輪機發電）、合成燃油 92，氫氣幾乎為零。先前進口 300–440 TWh 氫氣的結果，是因為氫氣是模型唯一可進口的載體；真正穩健的是進口能源的總量。
• Nuclear is not a question of cost here: even when allowed, the model builds none at these import prices or 25% lower. Only at 1.5 times the prices does it build 17 GW, which saves €2.5 bn a year (4%) and cuts imports from 278 to 171 TWh. The real trade-off is import dependence, which matters for a blockade.在此核電不是成本問題：即使允許興建，在這些進口價格或低 25% 時模型都不會興建。只有在價格為 1.5 倍時才興建 17 GW，每年節省 25 億歐元（4%），並將進口從 278 TWh 降至 171 TWh。真正的取捨在於進口依賴，這對封鎖情境至關重要。
• With Taiwan's own renewable costs (the feed-in tariff level: installed cost 1.0–2.7 times PyPSA's, higher O&M), 2050 costs €64.3 bn a year instead of €49.1 bn (+31%), and imports rise from 310 to 379 TWh. Allowed new nuclear now enters: 6.7 GW, saving €0.3 bn a year and cutting imports to 315 TWh (ammonia halves). Nuclear stays close to break-even, so the conclusion that it is mainly a question of import dependence holds.採用台灣本身的再生能源成本（躉購費率水準：設置成本為 PyPSA 的 1.0–2.7 倍，運維較高）時，2050 年系統成本為每年 643 億歐元，而非 491 億（+31%），進口由 310 TWh 增至 379 TWh。允許新核電時會興建 6.7 GW，每年節省 3 億歐元，進口降至 315 TWh（氨減半）。核電仍接近損益兩平，因此「核電主要是進口依賴問題」的結論不變。
• With 4-hour instead of daily time steps (which average out the solar day), the results hold: costs change by less than 1%, imports by 2–5%, and batteries grow from 5 to 33–50 GWh. Allowed nuclear falls from 6.7 to 5.5 GW, because batteries now carry solar power into the night.改以每 4 小時（而非每日，後者會平均掉太陽光電的日夜變化）為時間步長時，結果不變：成本變動小於 1%，進口變動 2–5%，電池由 5 GWh 增為 33–50 GWh。允許興建的核電由 6.7 GW 降為 5.5 GW，因為電池現在可將太陽光電移轉到夜間使用。
• Gas with carbon capture is built only when the official mix forces 20% thermal power; it would need part of the 40 Mt a year of CO₂ storage.只有在官方電力結構要求 20% 火力發電時才會興建燃氣碳捕捉電廠；它需要占用每年 40 Mt CO₂ 封存量的一部分。
• Still open: exposure of an import-based system to a blockade (designed, not yet built), the 2050 CO₂ storage volume, and import prices estimated for delivery to Germany rather than Taiwan.尚待處理：以進口為主的系統面對封鎖的風險（已完成設計，尚未建置）、2050 年 CO₂ 封存量，以及進口價格係以運至德國而非台灣估算。

What this draft shows此草稿顯示的結果

Assumptions and sources假設與來源

• Starting system: the official 2030 plan (MOEA supply-demand report 113年度: thermal schedule, Figure 3-3; renewable targets, Table 3-1).起始系統：官方 2030 年規劃（經濟部 113 年度全國電力資源供需報告：火力機組時程圖 3-3、再生能源目標表 3-1）。
• CO₂ cap: straight line from the model's 2025 reference (258.9 Mt) to net zero in 2050 (80% in 2030, 40% in 2040).CO₂ 上限：自模型 2025 年基準（258.9 Mt）直線降至 2050 年淨零（2030 年 80%、2040 年 40%）。
• Demand: electricity and the electronics industry +2.5%/yr (the official 2026–2035 outlook, carried to 2050); other demand flat. For comparison, the energy balance shows +1.8%/yr for household electricity in 2015–2025.需求：電力與電子業每年 +2.5%（官方 2026–2035 年展望，延用至 2050 年）；其他需求持平。對照：能源平衡表 2015–2025 年住宅用電年增 1.8%。
• CO₂ storage up to 40 Mt/yr, the negative-emissions volume in Taiwan's 2050 net-zero pathway (to verify).CO₂ 封存上限每年 40 Mt，取自台灣 2050 淨零路徑的負碳量（待查證）。
• Road transport: electric share of energy 5% (2030), 45% (2040), 85% (2050), derived from the EV action plan's targets of 30% electric new cars in 2030 and 100% electric new cars and scooters in 2040 (Table 3; the shares of energy are our derivation).公路運輸：電力占能源比例 5%（2030）、45%（2040）、85%（2050），由電動車行動計畫的目標推估：2030 年新車 30% 電動、2040 年新車與機車 100% 電動（表 3；能源比例為本研究推估）。
• What the model may build: solar (including rooftop), onshore and offshore wind, batteries, gas plants, transmission, and PyPSA-Earth's sector technologies (heat pumps, electrolysers, hydrogen storage and fuel cells, carbon capture, direct air capture, synthetic fuels). The official runs add gas with carbon capture, hydrogen and ammonia turbines and fuel imports, and allow no new nuclear (one run allows it to price the option). Only the early test and sensitivities A–D could build unlimited new nuclear at the Lungmen and Maanshan sites.模型可新建：太陽光電（含屋頂型）、陸域與離岸風電、電池、燃氣電廠、輸電線路，以及 PyPSA-Earth 的部門技術（熱泵、電解槽、儲氫與燃料電池、碳捕捉、直接空氣捕捉、合成燃料）。官方情境另加入燃氣碳捕捉、氫能與氨渦輪機及燃料進口，且不新建核電（其中一個情境允許核電，以評估其價值）。只有早期測試與敏感度 A–D 可於核四與核三廠址不設上限地新建核電。