公開資料から核融合の正味電力を再構成する

Meaningful comparison of published tokamak net-electric outputs requires a specified electrical boundary and time basis. This study asks which outputs can be reconstructed from public evidence without using the reported net value as a calculation input. A purposively selected corpus contains eighteen design claims in nine documents, with hypothetical scenarios and historical comparators analysed separately. Claims are classified by the strongest available accounting route, and source ambiguities are retained rather than silently resolved. Seven claims permit target-excluded arithmetic, including four in one ST-E1 article. The ST-E1 Case 1 recipe gives 277.944 against 278 MWe reported; STEP gives 149 MWe; ARC gives 189.3 against 190 MWe at a partial plant scope; and Yinsen Figure 70 gives 55.6 MWe. Yinsen’s different heating inputs prevent substitution of that figure for its Table 2 result. EU-DEMO HCPB and a STEP conference manuscript contain unresolved numerical differences between source elements. For K-DEMO, both reported targets lie inside an explicitly conditional heat-allocation set, but agreement obtained by calibrating a recovery coefficient does not identify the source model. The findings distinguish reproducible electrical subtraction, matched thermal allocation and the evidence needed to establish cycle-energy closure. Explicit source locations, worked calculations and an embedded executable check support inspection of the principal results. The study does not validate reactor designs, estimate population disclosure rates or establish the effectiveness of its extraction checklist.