Datasets:

THULab
/

oil028_sample

license: cc-by-nc-4.0
task_categories:
  - tabular-classification
  - tabular-regression
  - time-series-forecasting
language:
  - en
tags:
  - tsfile
  - timeseries
  - synthetic
  - oil-and-gas
  - midstream
  - downstream
  - storage-tanks
  - inventory-management
  - api-650
  - api-653
  - mass-balance
  - xpertsystems
  - time-series
  - format:tsfile
pretty_name: OIL-028 — Synthetic Tank Storage Dataset (Sample) (TsFile)
size_categories:
  - 100K<n<1M
configs:
  - config_name: default
    data_files:
      - split: inventory_levels
        path: inventory_levels.tsfile
      - split: optimization_labels
        path: optimization_labels.tsfile
      - split: scada_telemetry
        path: scada_telemetry.tsfile
      - split: tank_integrity
        path: tank_integrity.tsfile
      - split: tank_master
        path: tank_master.tsfile
      - split: tank_temperature_profiles
        path: tank_temperature_profiles.tsfile

OIL-028 — Synthetic Tank Storage Dataset (Sample) (TsFile)

Apache TsFile version of xpertsystems/oil028-sample.

Overview

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.

Tables: 6 TsFile table(s), 294,500 rows in total.
- inventory_levels: 84,000 rows, 2025-01-01 00:00:00 → 2025-01-07 23:00:00.499000 (UTC)
- optimization_labels: 500 rows, 1970-01-01 00:00:00 → 1970-01-01 00:00:00.499000 (UTC)
- scada_telemetry: 125,000 rows, 1970-01-01 00:00:00 → 1970-01-01 00:02:04.999000 (UTC)
- tank_integrity: 500 rows, 1970-01-01 00:00:00 → 1970-01-01 00:00:00.499000 (UTC)
- tank_master: 500 rows, 1970-01-01 00:00:00 → 1970-01-01 00:00:00.499000 (UTC)
- tank_temperature_profiles: 84,000 rows, 2025-01-01 00:00:00 → 2025-01-07 23:00:00.499000 (UTC)

Schema (TsFile structure)

`inventory_levels.tsfile`

Time (INT64, milliseconds) — the source timestamp column.
TAG: none — the table is stored as a single device.
FIELD (5): tank_id (string), inventory_bbl (double), fill_pct (double), inflow_bbl (double), outflow_bbl (double)

`optimization_labels.tsfile`

Time (INT64, milliseconds) — synthetic row index over the stable source order; this table has no clock column.
TAG: none — the table is stored as a single device.
FIELD (4): tank_id (string), optimization_score (double), utilization_grade (string), failure_risk (double)

`scada_telemetry.tsfile`

Time (INT64, milliseconds) — synthetic row index over the stable source order; this table has no clock column.
TAG: none — the table is stored as a single device.
FIELD (5): sensor_id (string), tank_id (string), signal_value (double), latency_ms (double), signal_quality (double)

`tank_integrity.tsfile`

Time (INT64, milliseconds) — synthetic row index over the stable source order; this table has no clock column.
TAG: none — the table is stored as a single device.
FIELD (5): tank_id (string), corrosion_rate_mpy (double), sludge_depth_ft (double), leak_probability (double), integrity_score (double)

`tank_master.tsfile`

Time (INT64, milliseconds) — synthetic row index over the stable source order; this table has no clock column.
TAG: none — the table is stored as a single device.
FIELD (7): tank_id (string), tank_type (string), product_type (string), capacity_bbl (int64), diameter_ft (double), height_ft (double), commission_year (int64)

`tank_temperature_profiles.tsfile`

Time (INT64, milliseconds) — the source timestamp column.
TAG: none — the table is stored as a single device.
FIELD (4): tank_id (string), temperature_f (double), ambient_temp_f (double), vapor_pressure_index (double)

Conversion notes

One TsFile per source CSV (6 tables) for this tank-farm inventory sample; each table is stored as a single device, so the rig/well/tank/sensor ids stay ordinary FIELDs.
Time = the table's own timestamp column as INT64 milliseconds (event timestamps, daily record/production dates, or test dates). Master/inventory tables that carry no clock (rigs_master, sensors_master, wells_master, tank_master, the label tables and scada_telemetry) use a synthetic row index as Time.
No column is dropped. Rows that share a timestamp are ordered by a 1 ms step in stable source order; daily tables keep their exact calendar day.

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("inventory_levels.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())

Source & license

Original dataset: xpertsystems/oil028-sample
Source card title / publisher: OIL-028 — Synthetic Tank Storage Dataset (Sample)
License: cc-by-nc-4.0 (as declared by the original dataset).