DOE Big Hill Sweet refinery assay validation
Issue #3305 requires refinery characterization to be qualified against public refinery data before atmospheric/vacuum fractionation and conversion-unit models are expanded. This page freezes the first external refinery-assay benchmark used by the campaign.
The later public DOE workbook, reported on 24 September 2021, is treated as a separate frozen source rather than silently replacing this 1998 MLI 009 matrix. Its per-cut UOP K values qualify the DOE Big Hill Watson-factor calculation.
Source and provenance
The benchmark uses the U.S. Department of Energy Strategic Petroleum Reserve (SPR) Big Hill Sweet, sample MLI 009, assay date 1998-05-04, published in Exhibit D of the SPR Standard Sales Provisions in the Code of Federal Regulations archive:
- GovInfo source: https://www.govinfo.gov/content/pkg/CFR-2004-title10-vol4/pdf/CFR-2004-title10-vol4-chapII-subchapI.pdf
- DOE SPR Crude Oil Assay Manual, 5th edition, 1 August 2024: https://www.spr.doe.gov/reports/docs/CrudeOilAssayManual.pdf
- DOE SPR assay landing page: https://www.spr.doe.gov/reports/Crude_Oil_Assays.html
The 2024 DOE manual documents atmospheric/light-vacuum fractionation by ASTM D2892 followed by D5236 for the residuum. It also states that distillation fractions are measured on a mass-percent basis and that volume-percent values are calculated using each fraction’s specific gravity. The paired volume/weight yields and density information therefore provide a transparent public-source consistency qualification for NeqSim’s volume-to-mass assay bookkeeping.
The volume and weight yields are not statistically independent measurements: DOE states that volume percentages are derived from measured mass percentages using fraction specific gravity. The regression is intentionally a bookkeeping/provenance qualification, not an independent validation of a petroleum-property correlation.
No ASTM equation or proprietary commercial-simulator implementation is reproduced. The regression freezes only numerical assay facts openly published by the U.S. Government and cites their public source.
Frozen validation slice
The first regression intentionally uses only the bounded 175-1050 degF distillate slice. Each included interval has finite lower/upper cut temperatures plus published volume yield, weight yield, specific gravity, and API gravity. The unbounded light and 1050 degF+ tails are excluded rather than assigning invented boiling boundaries.
| Cut range (degF) | Volume % of whole assay | Weight % of whole assay | SG 60/60 F | API gravity |
|---|---|---|---|---|
| 175-250 | 9.8 | 8.6 | 0.7815 | 49.6 |
| 250-375 | 15.4 | 15.2 | 0.8305 | 38.9 |
| 375-530 | 15.5 | 15.2 | 0.8623 | 32.6 |
| 530-650 | 10.8 | 11.1 | 0.9226 | 21.9 |
| 650-1050 | 27.8 | 30.3 | 0.9477 | 17.8 |
The same source reports whole-crude specific gravity 0.8451 and API gravity 35.9 degAPI. Those values provide an additional consistency check on the conventional relation API = 141.5 / SG - 131.5.
Acceptance contract
OilAssayCharacterisationDoeBigHillTest qualifies three separate things:
- Published SG/API pairs agree with the conventional API relation within 0.05 degAPI, consistent with the source’s one-decimal API reporting.
-
After normalizing the bounded 175-1050 degF slice on its own basis, NeqSim’s density-based conversion
\[w_i = \frac{v_i SG_i}{\sum_j v_j SG_j}\]must reproduce the normalized DOE weight-yield shape with maximum absolute deviation below 0.007 mass fraction (0.7 percentage points).
- Supplying the same cuts through the public API-gravity input path must reproduce the mass-fraction shape obtained from the four-decimal specific-gravity values within 5e-5 mass fraction. This tolerance reflects the DOE table’s one-decimal API rounding, not a thermodynamic-model uncertainty.
Each published API input must also round-trip exactly through the conventional dimensionless SG60/60 relation to 1e-12. Physical density is exposed separately as SG60/60 multiplied by 999.016 kg/m3. This prevents the reference water density from being applied twice when API gravity is used as an assay input.
For the frozen values, the largest pre-CI volume-to-mass deviation is approximately 0.00667 mass fraction. The largest specific-gravity versus rounded-API mass-shape deviation is approximately 4.27e-5 mass fraction. These tolerances are therefore data-agreement gates rather than machine-precision tolerances.
The test then applies the normalized slice through OilAssayCharacterisation, requires finite positive pseudo-component molar masses and mole amounts, and requires reconstructed assay mass closure to 1e-10.
What this validates
This benchmark provides public-source evidence for:
- refinery cut basis handling;
- SG/API consistency of the published source data;
- equivalence of the specific-gravity and API-gravity user input paths within source reporting precision;
- liquid-volume to mass-fraction conversion;
- preservation of bounded cut ranges;
- pseudo-component generation from a real refinery-assay slice;
- exact reconstructed mass closure after pseudo-component creation.
What this does not validate
This increment deliberately does not claim validation of:
- ASTM D86/D1160-to-TBP conversion;
- the molecular-weight, critical-property, acentric-factor, or EOS correlations for each pseudo-component;
- the light-ends or 1050 degF+ residue representation;
- atmospheric or vacuum column product yields;
- vapor-liquid equilibrium or flash-root changes;
- hydrotreating, reforming, FCC, hydrocracking, or other conversion-unit models.
Those remain separate #3305 quality gates. The next dependency-ready increment should extend this public-data foundation into a complete crude/pseudo-component slate and a reproducible atmospheric-fractionation case with mass/energy closure, product yields/boiling ranges, convergence/repeatability, and runtime evidence.