SCADA, EMS, DMS and operational technology

A useful measurement is more than a number: retain point identifier, value, unit, sign convention, source timestamp, receipt timestamp, quality flag and applicable revision. Source time says when the underlying observation occurred; receipt time says when your system obtained it. Synchronisation errors can make an apparently recent value misleading. Declare timezone and daylight-saving treatment rather than joining on ambiguous clock strings.

Our synthetic read-only exercise uses decision time 12:00:00 and a teaching freshness threshold of 60 seconds inclusive. Generator output is 48 MW at age 10 seconds with good quality; export is 46 MW at age 120 seconds with good quality; breaker position is closed at age 5 seconds but suspect; bus voltage is 132 kV at age 20 seconds with good quality. Only the first and fourth pass both screening conditions. Passing this screen is necessary here, not sufficient to authorise control.

Do not subtract the stale 46 MW export reading from the current 48 MW generator reading and report 2 MW of measured losses. The observations do not share a usable time basis, and their electrical boundaries also need checking. A new receipt timestamp cannot turn an old source observation into a fresh one. Record the reason for exclusion; do not silently replace suspect values with zero.

For this exercise only, assume exactly 60 scheduled observations per point in a one-hour reporting window. Of 60 expected records, 54 arrive; 6 of those 54 are stale or bad-quality. Completeness is 54/60 = 90%. Usable-record coverage is 48/60 = 80%. Quality among received records is 48/54 ≈ 88.9%. All three are valid but answer different questions; label the numerator, denominator and exclusions.

Communications availability, measurement freshness and analytical completeness are different service measures. A dashboard can be available throughout an hour while repeatedly displaying an old value. A low network delay does not establish sensor calibration. Conversely, event-driven or deadband reporting may legitimately leave a timestamp unchanged; use that system's update contract and heartbeat semantics before treating age alone as failure.

Define acceptance criteria before measuring performance: expected reporting behaviour, tolerance, quality states, late-arrival handling, duplicate and revision policy. Report critical breaker or protection-related point issues separately rather than burying them in an average across thousands of healthy analogue points. Connect this data contract to Topic 54's controlled pipeline and Topic 48's ownership and assurance.

A state estimate combines a network model with available measurements to infer a consistent operating state. It is not a second independent physical instrument. A wrong breaker topology, scaling error or correlated measurement error can compromise interpretation. A converged calculation is not proof that the underlying network model and all inputs are correct.

Compare like with like: an instantaneous MW observation, a dispatch target for a defined interval and a metered MWh quantity are not interchangeable. Integrating samples estimates energy only under declared assumptions about behaviour between samples, missing periods and revisions. AEMO's generation-data documentation even warns that a field called MWH_READING in one published product represents instantaneous MW; a column name is not a sufficient unit definition.

Investigate a mismatch in order: identify the device and electrical boundary; inspect timestamps, units, scaling and sign; check quality and topology; then consider physical performance. Keep candidate causes separate. Do not adjust raw observations merely to force agreement with a dispatch target. Preserve the original evidence and document any transformation so a second analyst can reproduce it.

• AEMO — generation and load data definitions

In a fictional tabletop, an authorised command is sent at 12:01:00 and a communications acknowledgement arrives at 12:01:01. The only displayed device status is timestamped 11:58:00. You cannot conclude the physical state changed. Record the command, acknowledgement and stale indication separately; seek approved independent confirmation through the responsible operator. This is an evidence exercise, not a switching procedure.

A defensive architecture separates read-only analytics from control authority. Map trust boundaries, permitted identities and the route for approved remote support. Use named responsibility, restricted privilege, monitored access and a tested recovery plan appropriate to the system. NIST's OT guidance is a reference for safety- and reliability-aware security design, not an Australian switching licence or a site-specific control procedure.

For a commissioning or recovery exercise, specify test prerequisites, expected state, success evidence, stopping criteria, rollback ownership and independent review. Do not infer failover success because a login page is reachable. Test the required function and its data quality in an authorised non-production environment. Production changes require the site's qualified people and approved change process.

• NIST SP 800-82 Rev. 3 — final OT security guide (2023)

Use AEMO's scheduling error report register to select an actual incorrect-SCADA-input report. Record event date and publication date separately. Read the report before describing its cause, affected interval or market impact. Our 12:00 dataset is invented and must not be presented as the Dederang incident or any other historical event.

Independent task: change the export observation to age 45 seconds with good quality and change the bus-voltage observation to bad quality. Apply the same screening rule, state which points are now usable and explain why the count alone conceals a different operational risk. In a second reporting window, 57 of 60 observations arrive and 3 received observations fail the screen. Calculate completeness and usable coverage without changing the denominator.

Submit a timeline, point-quality table and 200-word handover note. Separate observed facts, hypotheses, affected decisions, immediate read-only checks and actions requiring authorised operators. Include an owner and evidence requirement for closure. Review against four criteria: units and time reconcile; source records are traceable; uncertainty is explicit; no recommendation exceeds the learner's authority. Link this work to Topics 15, 18, 52, 54 and 57.

• AEMO — scheduling error reports, including incorrect SCADA inputs