Deferment Anomaly Detection Report

R0: Zero-Dfrm

Flags rows where all deferment volumes are effectively zero for the given WELL_TYPE, which usually indicates non-events or data ingestion errors. Used to prune junk data or tune preprocessing step.

R1: OvrdEvents

Identifies events where duration exceeds 90 days. Catches stale or forgotten deferment events that inflate deferment reporting. For ongoing events (with no END_DATE), duration is evaluated versus 'as-of-today'.

R2: DateLogic

Checks date integrity: START_DATE after END_DATE, PRODUCTION_DAY before START_DATE, or after END_DATE. These patterns typically come from parse errors, timezone drift, or manual entry mistakes.

R3: GasUnits

Verifies unit consistency between gas in Sm³ and MSCF. Near-zero values are skipped to avoid division noise. Large deviations suggest unit mix-ups or conversion errors upstream.

R4: Dfrm-WellType

Ensures reported deferment volumes are compatible with WELL_TYPE. E.g. WI well doesnt report Oil/Gas/Condensate deferment. Similarly, GP should not report Oil or Water volumes, and OP should not report Condensate or Water volumes.

R5: ProdMethod

Validates DETAILED_PROD_METHOD against WELL_TYPE policy: allowed or forbidden codes. For WI only prod_method = C is allowed, accordingly for OP (NF, BP, EP, GL) and for GP (NF), methods are allowed.

R6: GrpCons

Within multi-well events, attributes should align across wells. The rule highlights differences from the groups modal values for COMMENTS, DEFER_CATEGORY_IND, DEFER_ACTIVITY_OWNER, EVENT_TYPE, or ASSET_TYPE.

R7: FctyCons

Detects well-to-facility or well-to-train switches. Real reroutes are possible, but we flag those as its a rare case.

R8: GrpDateGaps

In long multi-well events, finds day-gaps where one well drops out while peers remain active, despite the well being active the day before and after. This pattern often signals intermittent logging or partial backfill rather than a true participation gap.

R9: EventVolConsistency

For full-deferment-day rows, for each well the rule compares reported deferment to the typical in-event value; rows exceeding 5pct are flagged (adjacent rows are shown in evidence for context).

R10: AssetType

Detects asset-type misclassification: a well-caused event incorrectly coded to equipment/facility (EQPM/FCTY) or a non-well-caused event incorrectly coded as WELL. Unmapped or TBC options are listed for follow-up, but are not classified as anomaly.

R11: VolAnomaly

Flags unusual GOR (OP wells) or CGR (GP wells) relative to a rolling historical median based on full-day deferments. Only the first crossing day is reported to avoid noisy streaks, and the evidence notes baseline sample size and quality.

R12: EventType vs Potential

Reconciles EVENT_TYPE against a pre-event 6m rolling potential (defined for full day deferments only). If observed deferment is ≥95pct of potential, the event should not be reported as CONSTRAINT; If observed deferment is <95pct of potential, the event should not be reported as DOWN. This can work more precisely if Well Test data fed into algorithm.

Quality metrics per field:.

Showing first 50 rows — anomalies detected in 347 rows.

Showing first 50 rows — anomalies detected in 72 rows.

Showing first 50 rows — anomalies detected in 146 rows.

Showing first 50 rows — anomalies detected in 5 rows.

Showing first 50 rows — anomalies detected in 30 rows.

Showing first 50 rows — anomalies detected in 9 rows.

Showing first 50 rows — anomalies detected in 37 rows.

Showing first 50 rows — anomalies detected in 27 rows.

Showing first 50 rows — anomalies detected in 445 rows.

Showing first 50 rows — anomalies detected in 66 rows.

Showing first 50 rows — anomalies detected in 154 rows.