Manufacturing survey kWh inside Industrial electricity allocation #90
Replies: 10 comments
|
M1. Industrial class size (unchanged from Discussion #88) What this is: How much generation the Industrial customer class gets in total — not how that total is split among factories. Implementation (intended): Do not change the Industrial class target from #88 D0 (class MWh from EIA Table 2.2, which is annual sales by customer class, including Direct Use). Residential, Commercial, Transportation, and Exports stay on electricity-bill shares inside their classes. The Industrial class is larger than manufacturing. EIA’s industrial sales include mining, construction, and agriculture. MECS only surveys factories (industry codes 31–33). So MECS cannot replace the class total. Using MECS’s 807 TWh (2022 purchases) as the Industrial target would shrink Industrial and push the missing MWh onto other classes, which #88 already forbade. Not chosen: Replacing the Industrial class total with MECS TWh; moving Direct Use out of Industrial. Resolution: Keep the #88 Industrial class total (sales + Direct Use). This work only changes weights inside that total. |
|
M2. Two groups inside Industrial: factories vs everyone else What this is: The Industrial class is not only factories. We split it into manufacturing plants (MECS covers them) and a residual (MECS does not). Implementation (intended): Manufacturing = Industrial purchasers that appear in the existing Cornerstone factory map (the same industry keys used for MECS fuel tables, with the Table 7.7 name overlay in M4). Residual = every other Industrial purchaser: farms, mines, construction, utilities, the generation industry buying its own power, Industrial final demand, and any future Industrial code not in the factory map. Do not define the residual as a fixed hand list of farms/mines/construction. That list is incomplete (utilities and Industrial final demand would be missed). Use it only as a check that those codes land in the residual. Not chosen: Treating all Industrial purchasers as manufacturing; putting utilities in the MECS pool. Resolution: Two groups. Factories get MECS purchased-kWh shares (M4) inside their pool. Everyone else in Industrial keeps electricity-bill shares inside the residual pool. |
|
M3. How large the factory pool is vs the rest What this is: After M2 names the two groups, we still need to decide how much of the Industrial class each group gets before MECS rearranges factories. Implementation (intended): Split the Industrial class total using electricity bills. If factories are 80% of Industrial electricity spending, they get 80% of Industrial generation MWh as a pool. MECS then only changes which factories get that pool. Not chosen: Using an AEO-style physical split (manufacturing ~71% of industrial kWh) for pool sizes. That would invent a physical series for farms/mines/construction that we do not have, and would change those sectors even though MECS does not cover them. Resolution: Pool sizes follow electricity bills. MECS rearranges manufacturing only. Residual Industrial keeps today’s dollar shares of the class. |
|
M4. Which factory kWh series: purchased electricity (MECS Table 7.7) What this is: Which MECS number ranks factories. Table 7.7 is purchased electricity by manufacturing industry. We use EIA’s Electricity total column (local utility plus other suppliers — marketers, independent generators, off-site cogeneration). About half of manufacturing purchases in 2022 are not from the local franchise utility; using only that column would throw half the kWh away. We do not sum the two supplier columns after filling withheld cells. In 2022 furniture’s total is withheld, but the local-utility column is published. Adding after a fill would invent a fake total and skip the leftover assignment in M7. Not chosen:
Mapping note: The factory list in Cornerstone was built for MECS fuel Table 3.1. Table 7.7 uses a few different industry codes (paper mills Resolution: Production weights = MECS Table 7.7 Electricity total (purchased kWh), mapped onto Cornerstone factories. Shares of the manufacturing pool (M3), not MECS TWh as the class total. |
|
M5. Direct Use stays mixed in; we do not yet use onsite-generation data What this is: Direct Use is electricity generated and used on site (not bought from the grid). Discussion #88 already put it in the Industrial class total (D0: Industrial includes Direct Use from EIA Table 2.2). This round does not assign that slice to the plants that actually generate on site. MECS Table 11.3 reports onsite generation by manufacturing industry (about 97% combined heat and power). Paper, chemicals, and refining dominate. Using it would give those plants more of the Industrial class, and would take Direct Use away from farms/mines/construction (who would otherwise keep a slice via bill shares). Why not this round: It is a second MECS extract, a second pool, and it raises the chance that combined-heat-and-power plants are assigned more generation MWh than their purchased-power bill can pay for. When that happens, #88 already caps generation dollars at the bill and moves leftover MWh to others in the same class (“water-fill”). Published mixed-units MWh then follow the capped dollars, so those plants would not actually keep the extra onsite kWh. Direct Use is also only about 12% of Industrial-plus-Direct-Use in recent years. Most of manufacturing’s Industrial MWh already follow Table 7.7 purchases (M4). Table 11.3 would only reassign that ~12% slice. Not chosen: Splitting Industrial into a purchase pool (Table 7.7) and a Direct Use pool (Table 11.3) in this round. Resolution: Direct Use rides along on purchased-electricity shares (factories) and electricity-bill shares (residual). Table 11.3 is a documented follow-on. |
|
M6. Which MECS year What this is: MECS is run every four years. We have 2018 and 2022. The model has a 2017 table set and a published later year (2024). Implementation (intended): Use the 2018 survey when building the 2017 tables. Use the 2022 survey at the published model year. Do not interpolate kWh between survey years (the mix is not a smooth annual series). Not chosen: Always 2022, including on the 2017 tables (that would put a 2022 factory mix on 2017 bills). Resolution: 2018 MECS on the 2017 chain; 2022 MECS on the 2024 model. |
|
M7. Withheld MECS cells What this is: EIA withholds some cells: Implementation (intended): On the Electricity total column only: treat Do not use the GHG helper that fills Not chosen: Zeroing furniture (that would spread ~1% of manufacturing kWh onto every other factory and give furniture plants zero generation MWh). Filling from the local-utility column only (2022 furniture would get 2.8 TWh and miss the rest). Ranking withheld industries by their electricity bills against the whole manufacturing pool (that can overstate furniture). Resolution: Tiny cells → 0. One withheld 3-digit industry gets the US leftover. Otherwise leftover must be ~0. |
|
M8. When several Cornerstone industries share one MECS row What this is: MECS sometimes publishes one kWh number for a group that Cornerstone splits into several industries. Implementation (intended): Split that kWh in proportion to those industries’ electricity bills (the same bills already used to allocate generation). If subtracting children from a parent would go negative, treat the leftover as zero. Not chosen: Equal split; split by gross output; loading a second BEA Use table just for this split. Resolution: Shared MECS rows split on electricity bills. |
|
M9. No new on/off switch What this is: Discussion #88 D9 (do not add extra on/off switches) already said the existing electricity-disaggregation flag runs the EIA-anchored method. This work stays behind that same flag. Implementation (intended): When electricity disaggregation is on, factory MWh follow MECS. There is no separate “MECS on/off” config. Diagnostics can still compute the old dollar split on the same bills as a comparison table. Emission-factor isolation vs “EIA-anchored but still dollar Industrial” is a saved snapshot of current production, taken before this change — same idea as the freeze used when #88 replaced the old 3-way split. Not chosen: A fourth YAML flag so “EIA-anchored + dollar Industrial” remains a live config after merge. Resolution: No new flag. Production is MECS inside manufacturing. Comparisons are diagnostics plus a pre-change snapshot. |
|
M10. Commercial building surveys (CBECS) stay out What this is: CBECS is EIA’s survey of commercial buildings by building type (office, warehouse, food sales, …). It is not a survey of IO industries. EIA says building type is not the same as industry code. Office buildings, for example, map to many NAICS. Bedrock already uses CBECS for land and water, not for electricity. There is no CBECS electricity extract on the production path. Not chosen: Using CBECS to split Commercial (or Industrial) electricity among IO sectors. Resolution: CBECS is out of production electricity allocation. Optional later analysis only. Also out of this round (for the record)
|
Uh oh!
There was an error while loading. Please reload this page.
This discussion lists the modeling decisions for using the EIA Manufacturing Energy Consumption Survey (MECS) to split generation inside the Industrial customer class in Cornerstone production.
It does not replace Discussion #88. #88 already decided how much generation each customer class gets, how generation vs transmission vs distribution dollars are split, and that we use one national generation price. #88 explicitly left “using manufacturing survey kWh inside Industrial� as later work. These decisions are that later work.
Code is not in this discussion. A separate implementation plan will cite these decisions. Please comment on the numbered posts below in order; earlier choices constrain later ones.
EIA tables used here (plain names)
When a post says “Table 2.2� it means EIA’s Electric Power Annual sales by customer class (Residential, Commercial, Industrial, Transportation), plus Direct Use (electricity generated and used on site, not bought from the grid).
When it says “Table 7.7� it means MECS purchased electricity by manufacturing industry — how many kilowatt-hours factories actually bought. We use EIA’s Electricity total column (power from the local utility and from other suppliers). We do not use only the local-utility column.
When it says “Table 11.3� it means MECS onsite generation by manufacturing industry (mostly combined heat and power). We are not using that table in this round.
When it says “Table 3.1� it means MECS net electricity counted as a fuel. We are not using that table for these weights.
When it says “Table 7.1� it means MECS average purchase prices. We are not using that table (Discussion #88 already dropped EIA retail prices as the link between dollars and MWh).
Guiding principles
Summary
“#88 production� means the live EIA-anchored path: class MWh from EIA sales, one national generation price, leftover bill → transmission and distribution. Inside each class, MWh still follow electricity dollars.
*, less than 0.5 million kWh) as zero. If a whole 3-digit industry is withheld, assign the leftover vs the US manufacturing total. In 2022 that leftover is furniture (~9.2 TWh).Unchanged from #88 (not reopened): one national generation price; leftover bill after generation is transmission and distribution; if assigned generation would cost more than the bill, cap generation at the bill and move leftover MWh to others in the same class; no new flags; Make, the electricity 3×3, and mixed-units conversion factors.
Independent comments below: M1–M10.
All reactions