Do not turn a data-centre capacity condition into measured green-energy share
Capacity, metered energy and a supply commitment are different measures with different field boundaries.
Similar percentages can answer different questions
Reading a data centre’s green-energy plan requires separating a share of capacity from a share of consumed energy. Identical numbers can have different measurement boundaries and periods. Field staff should be able to explain which meter record or commitment supports each one. Truthound can check declared categories and numeric bounds in the resulting table; it does not establish the facility’s actual environmental performance.
The reply describes an application condition
In a 7 October 2026 parliamentary answer, Singapore’s METI explained electricity and water planning for data centres. It said applicants in the second capacity-call exercise were required to power at least 50% of capacity from green sources. That is not a measurement that 50% of annual energy consumption had been achieved. Preserve the distinction between a requirement and operational evidence.
Separate meters from commitments
A proposed review table retains facility identity, IT capacity, metered energy, source-share basis and measurement period. Identify whether a percentage comes from a contractual commitment or observation. If the facility boundary changes, a similarly named indicator can cover a different scope. Hold a percentage the owner cannot explain instead of admitting it to an operating report solely because it is the latest available value.
Keep the unit counterexample visible
MW capacity and MWh consumption remain different physical quantities even in one numeric column. Mixing a 0–1 fraction with a 0–100 percentage can also hide a mistake inside an apparently plausible range. The operator chooses a representation first. Being within that range does not establish fulfilment of a supply commitment or an improvement in measured facility efficiency.
Check approved metric and unit combinations
Truthound Core Python SDK 3.3.4 (open-source SDK, Polars backend) documents ColumnPairInSetValidator for approved metric-unit pairs. BetweenValidator can check owner-defined bounds and boundary inclusion, with explicit missingness policy. Passing these rules does not establish meter accuracy, the nature of an electricity source or regulatory applicability. Those questions require separate technical and authoritative evidence, rather than a wider interpretation of a valid numeric field.
Existing GX Core range rules are another route
Pandera 0.35.0 (MIT open-source Python DataFrame validation library; pandas, Polars and other supported backends) custom checks are an alternative for owner-defined metric-unit combinations and range conditions. Retain a data-centre reporting contract that already manages the same units, bounds and null policy. Compare Truthound only for checking that condition on the same prepared table. Neither tool selects the correct boundary or establishes actual green-energy use, and documentation provides no measured superiority on this event’s data.
Start with what field records can explain
Trace one facility measure back to its metering boundary and commitment definition. Do not rank facilities while period or unit remains unresolved. Explain plans separately from observations. A sustainability discussion about data-centre growth needs a record of what the value actually measures before it needs another attractive percentage, so that participants can distinguish a condition imposed on a project from evidence delivered by its operation.
Conditions
Separate facility capacity, metered energy and source-share units and boundaries.
Alternatives
Pandera 0.35.0 custom checks can retain the same approved units, bounds and null policy.
Limitations
An application condition is not achieved performance; MW capacity is not an MWh consumption share.
Evidence
- Event checked
- Subject checked
- Modified
