| ---
|
| license: cc-by-nc-4.0
|
| task_categories:
|
| - tabular-classification
|
| - tabular-regression
|
| - time-series-forecasting
|
| language:
|
| - en
|
| tags:
|
| - synthetic
|
| - oil-and-gas
|
| - midstream
|
| - downstream
|
| - storage-tanks
|
| - inventory-management
|
| - api-650
|
| - api-653
|
| - mass-balance
|
| - xpertsystems
|
| pretty_name: "OIL-028 — Synthetic Tank Storage Dataset (Sample)"
|
| size_categories:
|
| - 100K<n<1M
|
| ---
|
|
|
| # OIL-028 — Synthetic Tank Storage Dataset (Sample)
|
|
|
| **SKU:** `OIL028-SAMPLE` · **Vertical:** Oil & Gas / Storage Terminals
|
| **License:** CC-BY-NC-4.0 (sample) · **Schema version:** `oil028.v1`
|
| **Sample version:** `1.0.0` · **Default seed:** `42`
|
|
|
| A free, schema-identical preview of XpertSystems.ai's enterprise tank
|
| storage dataset for **inventory forecasting, tank utilization optimization,
|
| overfill prevention, tank integrity ML, SCADA telemetry analytics, and
|
| storage terminal operations**. The sample covers **500 tanks**
|
| across **6 product types** (Crude Oil, Diesel, Gasoline, Jet
|
| Fuel, LNG, LPG) and **3 tank type categories** (Floating Roof,
|
| Fixed Roof, Cryogenic), with **294,500 rows** linked across **6
|
| tables** spanning **168 hours (1 week) of hourly time-series**.
|
|
|
| **OIL-028's distinctive feature is the mass-balance-coupled inventory
|
| accumulation** — hourly inventory evolves as `inventory_t+1 = clip(0,
|
| capacity, inventory_t + inflow_t - outflow_t)` per tank, producing genuine
|
| time-series dynamics with realistic boundary clipping at 0% and 100%
|
| capacity.
|
|
|
| ---
|
|
|
| ## What's in the box
|
|
|
| | File | Rows | Cols | Description |
|
| |---|---:|---:|---|
|
| | `tank_master.csv` | 500 | 7 | Tank catalog: 6 product types × 3 API 650/620 tank types × capacity (100K-1.5M bbl) + diameter (40-250 ft) + height (20-80 ft) + commission year (1980-2024) |
|
| | `inventory_levels.csv` | 84,000 | 6 | **Mass-balance-coupled hourly inventory**: inventory + fill % + inflow + outflow per API 2350 overfill protection |
|
| | `tank_temperature_profiles.csv` | 84,000 | 5 | Product temp + ambient temp + vapor pressure index per ASTM D323 RVP framework |
|
| | `scada_telemetry.csv` | 125,000 | 5 | Per-tank SCADA telemetry: signal value + latency + signal quality |
|
| | `tank_integrity.csv` | 500 | 5 | API 653 inspection metrics: corrosion rate + sludge depth + leak probability + integrity score |
|
| | `optimization_labels.csv` | 500 | 4 | ML labels: optimization score + 3-class utilization grade (A/B/C) + failure risk |
|
|
|
| Total: **294,500 rows** across 6 CSVs, ~14.5 MB on disk.
|
|
|
| ---
|
|
|
| ## Calibration: industry-anchored, honestly reported
|
|
|
| Validation uses a **10-metric scorecard** with targets sourced exclusively to
|
| **named industry standards**: **API 650** (Welded Tanks for Oil Storage),
|
| **API 620** (Large Low-Pressure Storage Tanks), **API 653** (Tank Inspection /
|
| Repair / Alteration / Reconstruction), **API 575** (Inspection of Atmospheric
|
| and Low-Pressure Storage Tanks), **API 2000** (Venting Atmospheric and Low-
|
| Pressure Storage Tanks), **API 2350** (Overfill Protection for Storage
|
| Tanks), **API 12F / 12B** (Small Tank Specifications), **API 580/581**
|
| (Risk-Based Inspection), **EPA AP-42** (Vapor Emissions from Storage Tanks),
|
| **EPA Subpart Kb** (VOC Standards for Storage Tanks), **NFPA 30** (Flammable
|
| and Combustible Liquids Code), **ASTM D323 / D5191** (Reid Vapor Pressure),
|
| **ASTM D86** (Distillation / Volatility), **PHMSA 49 CFR 195** (Hazardous
|
| Liquid Pipeline Storage), **EIA Petroleum Supply Annual** (Tank Farm
|
| Utilization Statistics), **OSHA 29 CFR 1910.106** (Flammable Liquids
|
| Storage).
|
|
|
| **Sample run** (seed `42`, n_tanks=500, hours=168):
|
|
|
| | # | Metric | Observed | Target | Tolerance | Status | Source |
|
| |---|---|---:|---:|---:|---|---|
|
| | 1 | avg capacity bbl | 811703.6460 | 800000.0 | ±200000.0 | ✓ PASS | API 650 (Welded Tanks for Oil Storage) + EIA Petroleum Supply Annual — mean tank capacity for mixed transmission tank farm portfolio (100K-1.5M bbl range; 800K bbl median for crude/refined product tank farms) |
|
| | 2 | avg diameter ft | 150.2266 | 145.0 | ±30.0 | ✓ PASS | API 650 + API 12F — typical tank diameter for storage tank portfolio (40-250 ft range per API 650 spec; 120-180 ft typical for large refinery/terminal tanks) |
|
| | 3 | avg height ft | 50.5175 | 50.0 | ±10.0 | ✓ PASS | API 650 height-to-diameter ratio constraints — typical tank height for atmospheric/cryogenic storage portfolio (20-80 ft range; 40-60 ft typical per API 650 design) |
|
| | 4 | avg fill pct | 60.5478 | 60.0 | ±10.0 | ✓ PASS | EIA Petroleum Supply Annual + API 2350 (Overfill Protection) — typical tank farm utilization (50-70% operational target; >85% triggers overfill risk; <40% indicates over-capacity) |
|
| | 5 | avg corrosion rate mpy | 4.2395 | 4.0 | ±2.0 | ✓ PASS | API 653 (Tank Inspection / Repair) + API 575 — typical tank shell + bottom corrosion rate (2-8 mpy normal range; >10 mpy triggers RBI high-risk; <2 mpy for well-maintained tanks) |
|
| | 6 | avg integrity score | 84.3915 | 85.0 | ±8.0 | ✓ PASS | API 653 + API 580/581 RBI scoring — typical integrity score for mid-life tank portfolio (70-100 scale per API 653 RBI methodology; 85 mid-range for routine maintenance schedule) |
|
| | 7 | avg commission year | 2001.9780 | 2002.0 | ±8.0 | ✓ PASS | PHMSA tank registry + EIA Petroleum Supply Annual — typical commission year for mid-life portfolio (1980-2024 range; 2002 mean reflects 22-year average age for transmission tank farms) |
|
| | 8 | net flow std bbl | 4888.3326 | 4900.0 | ±1200.0 | ✓ PASS | EIA Petroleum Supply Annual tank turnover statistics + API 650 design flow rates — typical net flow standard deviation for hourly tank operations (4000-6000 bbl/hr for large-tank crude/refined product turnover) |
|
| | 9 | product type diversity entropy | 0.9952 | 0.97 | ±0.04 | ✓ PASS | 6-class product type taxonomy per EIA / EPA Subpart Kb (Crude Oil, Diesel, Gasoline, Jet Fuel, LNG, LPG) — 6-class diversity benchmark for mixed refinery / terminal storage portfolio, normalized Shannon entropy |
|
| | 10 | tank type diversity entropy | 0.9984 | 0.97 | ±0.04 | ✓ PASS | 3-class tank type taxonomy per API 650 + API 620 (Floating Roof per API 650, Fixed Roof per API 650, Cryogenic per API 620) — 3-class diversity benchmark, normalized Shannon entropy |
|
|
|
| **Overall: 100.0/100 — Grade A+**
|
| (10 PASS · 0 MARGINAL · 0 FAIL of 10 metrics)
|
|
|
| ---
|
|
|
| ## Schema highlights
|
|
|
| **`tank_master.csv`** — 6-class product taxonomy × 3-class tank type
|
| taxonomy per **API 650 / API 620**:
|
|
|
| | Product Type | Typical Tank Type | Storage Notes |
|
| |---|---|---|
|
| | Crude Oil | Floating Roof (API 650) | Vapor space minimization, EPA Subpart Kb |
|
| | Diesel | Fixed Roof / Floating Roof (API 650) | Lower vapor pressure than gasoline |
|
| | Gasoline | Floating Roof (API 650) | High vapor pressure per ASTM D323 RVP |
|
| | Jet Fuel | Fixed Roof / Floating Roof (API 650) | Quality-conditioned per ASTM D1655 |
|
| | LNG | Cryogenic (API 620) | -260°F operation, boil-off management |
|
| | LPG | Sphere / Bullet (API 620) | Pressurized storage, propane/butane |
|
|
|
| **`inventory_levels.csv`** — **mass-balance-coupled accumulation** (the
|
| real physics in this SKU):
|
|
|
| > inventory_t+1 = max(0, min(capacity, inventory_t + inflow_t - outflow_t))
|
| > fill_pct = (inventory / capacity) × 100
|
| > inflow ~ U(0, 12000) bbl/hr
|
| > outflow ~ U(0, 12000) bbl/hr
|
|
|
| This generates genuine time-series with **boundary clipping at 0% (empty)
|
| and 100% (overflow)** per API 2350 overfill protection. The sample's net
|
| flow standard deviation is ≈ 4900 bbl/hr — bullseye for **EIA Petroleum
|
| Supply Annual** tank turnover statistics for large tanks.
|
|
|
| **`tank_integrity.csv`** — **API 653 tank inspection metrics**:
|
|
|
| > corrosion_rate ~ U(0.5, 8.0) mpy (API 575 / NACE typical range)
|
| > sludge_depth ~ U(0, 5) ft (API 653 bottom inspection)
|
| > leak_probability ~ U(0.0001, 0.05) (PHMSA tank incident statistics)
|
| > integrity_score ~ U(70, 100) (API 653 RBI scoring)
|
|
|
| ---
|
|
|
| ## Suggested use cases
|
|
|
| 1. **Inventory forecasting** — time-series forecasting of `inventory_bbl`
|
| from inflow + outflow features per mass balance physics. **Strong
|
| physics signal** — within-tank dynamics are deterministic.
|
| 2. **Fill % boundary detection** — predict overfill (fill_pct > 90%) or
|
| underutilization (fill_pct < 30%) per API 2350 overfill protection.
|
| 3. **Net flow regression** — predict `inflow - outflow` net flow per
|
| tank from time + capacity features.
|
| 4. **Tank turnover analytics** — compute tank turnover frequency per
|
| tank-week from inventory time-series.
|
| 5. **3-class tank type classification** — predict `tank_type` from
|
| capacity + diameter + height features per API 650/620 design.
|
| 6. **6-class product type classification** — predict `product_type`
|
| from tank physical features. (Note: in sample, product-tank type
|
| coupling is not enforced — see Honest Disclosure §1.)
|
| 7. **Corrosion rate regression** — regression on `corrosion_rate_mpy`
|
| per API 653 inspection baselines.
|
| 8. **API 2350 overfill alerting** — anomaly detection on inventory
|
| approaching 95%+ capacity.
|
| 9. **Tank utilization optimization** — 3-class utilization grade
|
| classifier (rare-feature-coupled at sample scale; see Honest
|
| Disclosure §6).
|
| 10. **Multi-table relational ML** — entity-resolution learning across
|
| the 6 tables via `tank_id`. Mass-balance inventory + integrity
|
| metrics + utilization labels form joinable training pipelines.
|
|
|
| ---
|
|
|
| ## Loading
|
|
|
| ```python
|
| from datasets import load_dataset
|
| ds = load_dataset("xpertsystems/oil028-sample", data_files="inventory_levels.csv")
|
| print(ds["train"][0])
|
| ```
|
|
|
| Or with pandas:
|
|
|
| ```python
|
| import pandas as pd
|
| tanks = pd.read_csv("hf://datasets/xpertsystems/oil028-sample/tank_master.csv")
|
| inv = pd.read_csv("hf://datasets/xpertsystems/oil028-sample/inventory_levels.csv")
|
| temp = pd.read_csv("hf://datasets/xpertsystems/oil028-sample/tank_temperature_profiles.csv")
|
| integ = pd.read_csv("hf://datasets/xpertsystems/oil028-sample/tank_integrity.csv")
|
| labels = pd.read_csv("hf://datasets/xpertsystems/oil028-sample/optimization_labels.csv")
|
|
|
| # Mass-balance feature engineering for inventory ML:
|
| inv['net_flow'] = inv['inflow_bbl'] - inv['outflow_bbl']
|
| inv_avg = inv.groupby('tank_id').agg(
|
| avg_inventory=('inventory_bbl', 'mean'),
|
| avg_fill_pct=('fill_pct', 'mean'),
|
| net_flow_std=('net_flow', 'std')
|
| ).reset_index()
|
|
|
| # Multi-table join:
|
| joined = (tanks
|
| .merge(inv_avg, on='tank_id')
|
| .merge(integ, on='tank_id')
|
| .merge(labels, on='tank_id'))
|
| ```
|
|
|
| ---
|
|
|
| ## Reproducibility
|
|
|
| All generation is deterministic via the integer `seed` parameter (driving
|
| `np.random.seed`). A seed sweep across `[42, 7, 123, 2024, 99, 1]`
|
| confirms Grade A+ on every seed in this sample.
|
|
|
| ---
|
|
|
| ## Honest disclosure of sample-scale limitations
|
|
|
| This is a **sample** product calibrated for tank storage ML research, not
|
| for live operational decisions. **Several important limitations should be
|
| understood before use:**
|
|
|
| 1. **Product type does NOT drive tank type assignment.** The sample's
|
| product-tank type cross-tabulation is approximately uniform —
|
| **LNG can appear in Fixed Roof tanks** (physically impossible —
|
| LNG mandates Cryogenic API 620), **LPG can appear in Floating Roof**
|
| (impossible — LPG mandates Sphere/Bullet API 620), and **Gasoline
|
| appears in Fixed Roof** (uncommon — gasoline mandates Floating Roof
|
| per EPA Subpart Kb). Use the full product v1.1 for physics-mandated
|
| product-tank type coupling, or **filter the sample to physically-
|
| valid combinations** before training:
|
| ```python
|
| valid = ~((tanks['product_type'] == 'LNG') & (tanks['tank_type'] != 'Cryogenic'))
|
| tanks_valid = tanks[valid]
|
| ```
|
|
|
| 2. **Product temperature does NOT track ambient.** The sample's product
|
| temp ↔ ambient correlation is r ≈ 0.001 (essentially independent).
|
| Real tank temperatures track ambient with thermal lag per Newton's
|
| cooling. For thermal-conditioned ML, **use the temperature field as
|
| a noisy random feature**, not a true ambient-coupled signal.
|
|
|
| 3. **Vapor pressure index is NOT product-coupled.** The
|
| `vapor_pressure_index` field is uniformly sampled `U(0.5, 1.5)`
|
| regardless of product type. Real RVP depends on product (gasoline
|
| ~9 psi summer / ~15 psi winter, diesel ~0.04 psi, LPG ~120 psi per
|
| ASTM D323) and temperature (Antoine equation). For VOC emissions ML
|
| per EPA AP-42, **derive your own product-conditioned RVP** from
|
| industry tables.
|
|
|
| 4. **SCADA signal is uniformly random.** The `scada_telemetry.signal_value`
|
| field is sampled `U(0, 100)` without coupling to inventory, temperature,
|
| or tank physics. For SCADA-specific ML, use **OIL-024 SCADA telemetry**
|
| (which has signal-type-conditioned latency) or **OIL-025 acoustic
|
| telemetry** (which has physics-coupled signal generation).
|
|
|
| 5. **Integrity score is NOT corrosion-coupled.** The sample's corrosion
|
| rate ↔ integrity score correlation is r ≈ 0.01 (essentially independent).
|
| Real API 653 RBI integrity scoring couples to corrosion rate and
|
| wall loss. For integrity ML, **derive your own integrity scores**:
|
| ```python
|
| integ['derived_integrity'] = 100 - integ['corrosion_rate_mpy'] * 5 - integ['leak_probability'] * 100
|
| ```
|
|
|
| 6. **All optimization labels are random.** The 3-class `utilization_grade`
|
| (A/B/C) and the continuous `optimization_score` and `failure_risk`
|
| are sampled independently of any tank feature. **Sample's
|
| utilization_grade↔fill_pct correlation is ~ noise.** For utilization
|
| ML, **derive your own labels from physics features**:
|
| ```python
|
| labels['derived_grade'] = pd.cut(inv_avg['avg_fill_pct'],
|
| bins=[0, 40, 75, 100],
|
| labels=['C', 'B', 'A'])
|
| ```
|
|
|
| 7. **Leak probability is independent of corrosion + age.** The
|
| `leak_probability` field is sampled `U(0.0001, 0.05)` without
|
| coupling to corrosion_rate, commission_year, or product type. Real
|
| tank leak rates couple to NACE corrosion + age + product chemistry.
|
|
|
| 8. **Inflow and outflow are independent.** Real tank operations have
|
| correlated I/O cycles (filling vs draining phases), but the sample
|
| samples both uniformly per hour. **For tank turnover ML, expect the
|
| mean net flow to be near zero** (≈ -2.3 bbl/hr in sample), reflecting
|
| uncorrelated I/O cancellation rather than realistic operational
|
| cycles.
|
|
|
| ---
|
|
|
| ## Where physics IS strong (use these for ML)
|
|
|
| Three coupling signals in this sample are **physically valid and ML-useful**:
|
|
|
| | Signal | Physics | Use For |
|
| |---|---|---|
|
| | **Mass-balance inventory accumulation** | inventory_t+1 = clip(0, capacity, inventory_t + inflow - outflow) | Inventory forecasting, time-series ML |
|
| | **Fill % boundary clipping** | Realistic 0%/100% capacity bounds | API 2350 overfill detection |
|
| | **Tank physical dimensions** | API 650 capacity ↔ diameter ↔ height geometry | Tank classification, design ML |
|
|
|
| ---
|
|
|
| ## Cross-references to other XpertSystems OIL SKUs
|
|
|
| This SKU is the **first tank storage SKU** in the catalog — complementing
|
| the midstream pipeline operations stack with **storage terminal physics**:
|
|
|
| | SKU | Layer | Focus |
|
| |---|---|---|
|
| | OIL-015 | Midstream pipeline | Flow assurance (wax / hydrate threshold gating) |
|
| | OIL-024 | Midstream pipeline | Hydraulics + SCADA + 15 transient events |
|
| | OIL-025 | Midstream pipeline | Leak detection + acoustic + RBI |
|
| | OIL-027 | Midstream pipeline | Corrosion progression + cathodic protection |
|
| | **OIL-028** | **Storage terminal** | **Tank storage + mass-balance inventory + API 650/653** *(this SKU)* |
|
|
|
| **OIL-028 vs OIL-024/025**: OIL-024/025 simulate **flowing pipeline
|
| operations**. OIL-028 simulates **terminal tank operations** at pipeline
|
| endpoints (refinery tank farms, marine terminals, distribution depots).
|
| Use OIL-024/025 for **pipeline ML**, OIL-028 for **terminal/tank ML**.
|
|
|
| **OIL-028 vs OIL-019/020**: OIL-019/020 simulate **refinery process
|
| operations** (CDU/VDU/FCC + product yields). OIL-028 simulates **storage
|
| operations** that feed/receive from refining. Use OIL-019/020 for
|
| **refining ML**, OIL-028 for **storage logistics ML**.
|
|
|
| ---
|
|
|
| ## Full product
|
|
|
| The **full OIL-028 dataset** ships at **5,000 tanks × 8,760 hours (1 year)**
|
| (prod mode) producing tens of millions of rows with **physics-mandated
|
| product-tank type coupling per API 650/620**, **product-conditioned vapor
|
| pressure per ASTM D323 + Antoine equation**, **correlated inflow/outflow
|
| operational cycles per refinery scheduling**, **feature-coupled integrity
|
| scoring per API 653 RBI**, **age-coupled leak probability per NACE +
|
| PHMSA**, and **ambient-coupled thermal dynamics per Newton's cooling** —
|
| licensed commercially. Contact XpertSystems.ai for licensing terms.
|
|
|
| 📧 **pradeep@xpertsystems.ai**
|
| 🌐 **https://xpertsystems.ai**
|
|
|
| ---
|
|
|
| ## Citation
|
|
|
| ```bibtex
|
| @dataset{xpertsystems_oil028_sample_2026,
|
| title = {OIL-028: Synthetic Tank Storage Dataset (Sample)},
|
| author = {XpertSystems.ai},
|
| year = {2026},
|
| url = {https://huggingface.co/datasets/xpertsystems/oil028-sample}
|
| }
|
| ```
|
|
|
| ## Generation details
|
|
|
| - Sample version : 1.0.0
|
| - Random seed : 42
|
| - Generated : 2026-05-23 00:24:28 UTC
|
| - Tanks : 500
|
| - Simulation hours : 168 (1 week hourly)
|
| - Product types : 6 (Crude Oil, Diesel, Gasoline, Jet Fuel,
|
| LNG, LPG)
|
| - Tank types : 3 (Floating Roof per API 650, Fixed
|
| Roof per API 650, Cryogenic per API 620)
|
| - Capacity range : 100,000 - 1,500,000 bbl (API 650 portfolio)
|
| - Diameter range : 40 - 250 ft (API 650 spec)
|
| - Height range : 20 - 80 ft (API 650 spec)
|
| - Commission years : 1980 - 2024
|
| - Calibration basis : API 650, API 620, API 653, API 575, API 2000,
|
| API 2350, API 12F/12B, API 580/581, EPA AP-42,
|
| EPA Subpart Kb, NFPA 30, ASTM D323/D5191, ASTM D86,
|
| PHMSA 49 CFR 195, EIA Petroleum Supply Annual,
|
| OSHA 29 CFR 1910.106
|
| - Overall validation: 100.0/100 — Grade A+
|
| |
