Technology / Engineering guide
Wind Data Quality Flags & Aggregation Rules
Quality status describes how a record can be used, separately from its numerical value. A zero, a missing value and an estimated replacement are different information. Preserve those distinctions when calculating means, energy totals or coverage.
Carry record status into the calculated result.
Aggregation should explain which observations contributed and what part of the intended period they represent.
01 / Value and status
Keep the measurement separate from its eligibility.
A valid zero can describe an observed quantity of zero.
A valid zero can describe an observed quantity of zero. A missing record says that the quantity was unavailable. An invalid record may contain a number but lack evidence for its use. These cases should not be mapped to one generic zero value.
The CF conventions illustrate self-describing flags with stated meanings. They are a metadata example, not evidence of a MARSTON flag implementation. The actual data source or analyst must define any statuses used in a project.
02 / Flag dictionary
Describe each meaning and its origin.
A useful flag dictionary records code or label, meaning, source and applicable variable.
A useful flag dictionary records code or label, meaning, source and applicable variable. Distinguish a device-reported status from a later analyst assessment. If several conditions can coexist, retain how those combinations are represented.
Avoid inventing a universal “good” code shared by all devices. An undocumented code is unresolved information, not permission to infer a valid observation. Record revision changes when the dictionary or processing rule changes.
03 / Inclusion rule
Match eligibility to the analysis purpose.
A record unsuitable for performance comparison may still be useful for documenting an outage or reviewing data delivery.
A record unsuitable for performance comparison may still be useful for documenting an outage or reviewing data delivery. Define inclusion separately for each intended result rather than deleting every flagged record from the master dataset.
Retain raw values and statuses while producing an analysis view. This makes a later decision about eligibility reversible and explainable. The source-quality guide addresses checks on the original resource files.
04 / Aggregation basis
Keep the denominator visible.
For a mean, identify the contributing durations or samples and their weighting.
For a mean, identify the contributing durations or samples and their weighting. For an energy sum, missing intervals leave unobserved energy rather than proving no production. A partial sum should stay labelled with its represented coverage.
In a hypothetical set of four equal intervals, valid values 0, 10 and 20 plus one missing value have an observed mean of 10 over three intervals. Replacing the missing value with zero creates 7.5 over four; that is a different assumption, not recovered information.
05 / Estimated values
Show replacement assumptions separately.
If an assessment estimates missing values, identify the method, affected intervals and uncertainty.
If an assessment estimates missing values, identify the method, affected intervals and uncertainty. An estimated value should not become indistinguishable from an observation because both appear in the same output column.
Report observed contribution and estimated contribution separately where they influence a decision. Different replacement methods can change a total even when the raw records are unchanged. Retain enough provenance to reproduce that difference.
06 / Result metadata
Make coverage part of the delivered statistic.
A useful result includes intended period, contributing coverage, exclusions, estimated portions and processing version.
A useful result includes intended period, contributing coverage, exclusions, estimated portions and processing version. This prevents a valid mean over available records from being presented as a complete-period result.
The availability guide separates data coverage from equipment availability. No product data flags or completeness metric are confirmed here; the assessment should use the actual supplied records and documented status meanings.
Continue the review with Wind Monitoring Timestamps, Time Zones & Intervals.
Continue the review with Wind Energy Counters, Resets & Interval Totals.
Continue the review with Seasonal Wind Energy & Load Matching.
General background: CF conventions: self-describing flag values and meanings. These references explain industry concepts; they are separate from the applicable MARSTON product and project documents.
Prepare the information for this review.
Use the checklist to organise the project information. Mark unavailable items and retain document references.
| Review item | Information to provide | Purpose |
|---|---|---|
| Status meanings | Codes or labels, source and applicable variables | Interpret eligibility correctly. |
| Purpose rule | Which statuses contribute to each result | Avoid indiscriminate deletion. |
| Denominator | Intended period, included samples and duration weights | Explain the statistic. |
| Estimated portion | Affected intervals, method and uncertainty | Separate reconstruction from observation. |
| Result provenance | Coverage, exclusions and processing revision | Make the result reproducible. |
Engineering questions.
Can missing values be treated as zero?
Only as an explicitly justified modelling assumption; they are not observed zeros.
Does every flag mean a record must be deleted?
No. Its eligibility depends on the result being calculated.
Do all controllers use the same flag codes?
No common MARSTON implementation is confirmed, and source dictionaries can differ.
Does a partial energy sum describe the complete period?
Only if the unobserved portion is separately addressed and the result labelled appropriately.
Is an estimated replacement a measured observation?
No. Preserve its method and status.
What should travel with an aggregated value?
Period, coverage, inclusion rule, weighting, estimated portions and processing identity.
Technical enquiry
Bring your project information to the discussion.
Quality status describes how a record can be used, separately from its numerical value. Share the available information and the equipment scope you want reviewed in a technical quotation.