Energy Flow Diagrams — Sweden & Japan

Sources

• METI / Agency for Natural Resources and Energy, 総合エネルギー統計 — Comprehensive Energy Statistics, every fiscal year from FY1990 to FY2024 (確報, revised), energy-unit sheet, terajoules (stte_<FY>.xlsx). METI restates the whole back series on the current methodology, so all 35 years are on one basis.
• FY2023 press release of 25 April 2025: 令和5年度エネルギー需給実績（確報）.
• Japanese fiscal years run April–March; FY2024 covers April 2024 – March 2025.

Verified totals — reference years

All 35 fiscal years are in the workbook; three are shown here as anchors.

The FY2023 figures match METI’s published headline (17,558 PJ supply, 11,509 PJ final consumption). Every conversion box closes exactly in all 35 years. FY2012–FY2014 have almost no nuclear at all: every Japanese reactor was offline through FY2014, so those charts show the energy system as it stood at the bottom of the post-Fukushima shutdown.

How the numbers were derived

• Sign convention: in the METI table, transformation rows are negative for fuel consumed and positive for energy produced. Primary supply is row #190000 (国内供給), already net of exports.
• Aggregated bands net out product manufacture inside the band. Coal covers coal and coal products, oil covers crude and oil products, gas covers natural gas and city gas. Because city gas is manufactured from natural gas (plus about 96 PJ of LPG feedstock from the oil band), the rows for coke ovens, refineries, gas works and inter-product transfers (#210000, #220000, #230000, #280000) are netted into each band — otherwise the natural gas that becomes city gas would be counted twice. Stock change (#350000) is included as supply, since a stock drawdown is energy supplied to the system that year.
• Electricity generation merges utility power (事業用発電, including pumped-storage losses) and auto-producers (自家用発電, which contains most on-site and rooftop solar PV).
• Steam & heat supply merges industrial auto-steam boilers (自家用蒸気) with the district heat-supply business (熱供給); most of its output returns to manufacturing.
• Sectors: Industrial = 農林水産鉱建設業 + 製造業; Commercial & services = 業務他; each is shown net of the non-energy use booked inside it, which is drawn as its own terminal.
• Waste & recovered energy is 未活用エネルギー (refuse-fired power, waste tyres and plastics, recovered industrial heat).
• Renewables are split using METI’s $N detail columns: solar PV and solar heat, wind, biomass, geothermal. Nuclear, hydro, geothermal and PV use METI’s primary-equivalent accounting (thermal input for nuclear and geothermal; generated electricity for hydro, PV and wind).
• Own use & losses is the residual of each band after conversion, final and non-energy use, so it bundles transformation losses, energy-industry own use, transmission losses and the statistical difference — matching how the LLNL charts present these.
• End use applies the LLNL efficiency assumptions: residential and commercial 65%, industrial 49%, transport 21%.

Primary-energy convention (important when reading band widths)

METI uses the substitution method: every electricity-generating primary source is converted to primary-energy equivalent at a single reference efficiency. Recovered from METI’s own natural-units sheet (generation in GWh × 3.6 ÷ the primary energy booked):

The factor is not fixed across the series — it rises from 38.66% in FY1990 to 42.86% in FY2014 and 42.41% in FY2024, and is recovered from each year’s own natural-units sheet rather than assumed. Within a year it is applied identically to every generating source present. Their band widths are therefore mutually comparable, but each is about 2.4× the electricity actually generated. Geothermal is the exception, reported directly as heat. Rejected energy leaving the electricity node consequently includes this accounting artefact rather than only physical losses. Sweden uses a different convention (nuclear as reactor heat at 36%, renewables as electricity at 100%), so band widths must not be compared between the two charts.

Because band widths represent throughput including stock change and intra-band transformation, the bands sum to slightly more than METI’s domestic-supply headline (17,629 vs. 17,558 PJ in FY2023, a 0.4% difference). The headline figure above is METI’s own and is the one to cite.

Comparison with the IEA energy Sankey

For 2023 the IEA reports a total energy supply of 15,844 PJ (calendar year) against METI’s 17,558 PJ (fiscal year). The ≈1,714 PJ gap is fully explained by four convention differences — neither figure is wrong:

• Calorific basis (≈ +970 PJ): METI books fuels at gross calorific value (HHV), the IEA at net (LHV). This raises gas by about 354 PJ, oil 409 PJ and coal 204 PJ. Japanese national statistics always sit a few percent above IEA figures for this reason.
• Renewable electricity accounting (≈ +916 PJ): METI converts hydro, solar and wind electricity to primary equivalent by dividing by average fossil-plant efficiency (implied factor 41.8%: IEA hydro 271 PJ of actual electricity vs. 648 PJ here). The IEA counts these 1:1. Partly offset by nuclear (METI actual ≈38.5% heat, 724 PJ, vs. IEA fixed 33%, 917 PJ: −193 PJ) and geothermal (IEA imputes 10% efficiency: −93 PJ).
• Recovered energy (≈ +273 PJ): METI counts recovered industrial waste heat and electricity (未活用エネルギー) as primary supply; the IEA does not.
• Fiscal vs. calendar year plus LHV conversion of biomass and waste: the remainder (≈ −150 PJ).

A consequence of the substitution method: roughly 0.9 EJ of the rejected energy leaving the electricity box is an accounting artefact for hydro, solar and wind rather than physical waste heat. LLNL uses the same convention in its U.S. chart.