The grid went dark. The explanation took months.

What happened

ENTSO-E records a total blackout in continental Spain and Portugal on 28 April 2025. The rest of the European system had no significant disturbance. Its final report, published on 20 March 2026, identifies interacting technical factors rather than one simple cause.

A national trend is the start of a question

Imagine a control room comparing a public load series with an incident reported by a site team. The chart provides context for the wider system. The local report describes one observation at one place. Even when their times appear close, that does not establish that one caused the other. The analyst must know the geographic scope, units and meaning of each record before using them together. This is an illustrative workflow, not a reconstruction of an ELIA incident.

The practical value of connecting sources is that a reviewer can ask a better question: does an authorised asset record or a verified site report explain the observation? If the information is unavailable, that gap stays visible. A forecast that has been revised should not replace a historical measurement, and a national figure should not become the status of a local asset. A useful operational workspace makes the limits of the available evidence easy to inspect.

One event, different clocks

A grid event, an equipment record, a field observation and a control-room decision describe different parts of the same situation. Before they can support a reliable reconstruction, their times, units and source identifiers must line up. A national trend cannot establish what happened at a particular substation.

That is the operational lesson we draw from the investigation, not a finding that ELIA lacked information or a claim that a different platform would have prevented the blackout. Belgium was not part of the Iberian outage.

What this means in a Belgian control room

ELIA publishes historical total-load data in ods001 and near-real-time wind estimates and forecasts in ods086. These are useful contextual sources. They are not a substitute for authorised operational telemetry, asset records or observations from a site team.

• Keep the source visible. Show whether a value is a measurement, estimate or forecast, with its observation time and ingestion time.
• Connect the right asset. Use approved identifiers to relate an incident, a site and its responsible team. Do not invent a site-level state from aggregate public data.
• Make uncertainty visible. Flag delayed or missing feeds and keep corrections traceable. A missing value is not zero.
• Respect the control boundary. Read approved OT exports through the existing DMZ or data diode. No direct controller polling and no write-back from the incident workspace.

With one shared picture

Imagine an illustrative substation incident: the field team confirms a local observation while the crisis room sees the same asset, report and decision timeline. Public grid data supplies wider context. Engineers keep responsibility for electrical diagnosis, protection settings and switching; the shared picture supports coordination.

The ontology connects objects and keeps their meaning

A site, a grid asset, an organisation and a field team have different identities and responsibilities. A shared model can relate them without treating similar names as proof that they are the same thing. Before a connector is accepted, users need an agreed mapping between the source identifiers and those operational objects. The source record remains available alongside the interpretation, with its own timestamp and history.

For example, a team may submit a report against a recognised site while an imported event refers to an equipment identifier. An approved mapping can place both in the same incident context. If that mapping is uncertain or no longer current, the workspace should ask for review rather than quietly join the records. Local AI could assist with summarising the reports, but it should not invent missing telemetry, diagnose the electrical cause or authorise an intervention. Those responsibilities stay with qualified people and their existing systems.

Evaluate one connection before extending the picture

A first Meridian integration could use selected public datasets, an approved asset sample and example field reports. The operator and engineering team would define the read boundary, permitted users and expected update behaviour. An exercise could delay one feed, correct an identifier and introduce a contradictory report. Reviewers should still be able to find the original observation and explain why the incident view changed.

Useful acceptance questions are concrete: can the team distinguish stale data from a genuine change, identify the responsible source and follow an assigned verification task? The pilot must also test the behaviour of a failed connector without affecting the operational source system. European control means knowing where selected records and AI processing reside, who maintains the integration and how access is enforced. The result should support existing responsibilities while keeping electrical control and independent continuity arrangements separate.

This is a proposed integration scenario, not an ELIA deployment or partnership. A shared operational picture helps people compare evidence; it does not operate the electricity grid.