Fault DetectionStarter guide

How to separate PV equipment faults from telemetry gaps

A diagnostic workflow for distinguishing probable physical underperformance from communications and measurement problems.

ArraySignal Editorial Team7 min read

A zero or low production value does not by itself identify an equipment failure. The same pattern can be produced by a disconnected logger, delayed data, a frozen sensor, curtailment, nighttime operation, or a true electrical outage. A reliable diagnostic workflow separates data validity from physical interpretation before assigning a probable cause.

1. Start with timestamp integrity

Confirm that timestamps are timezone-aware, ordered, and unique at the intended measurement boundary. Measure the nominal cadence, identify gaps, and retain duplicate records as explicit quality issues until a deterministic resolution rule is applied. A site can appear to recover or fail simply because two systems report in different timezones.

2. Test signal plausibility

  • Flag negative AC production unless the measurement boundary legitimately includes import.
  • Flag sustained values above a documented capacity tolerance.
  • Detect frozen sequences while irradiance or peer-inverter signals continue to change.
  • Treat missing actual power as unknown rather than zero.
  • Exclude nighttime zero production from outage classification.

3. Seek independent evidence

Component-relative analysis is often the fastest discriminator. If one inverter falls while peer equipment and irradiance remain coherent, the evidence favors localized underperformance. If every measured channel freezes at the same timestamp, a communications problem is more likely. Status codes, plant-meter data, weather inputs, and nearby component behavior should corroborate rather than merely repeat the same source.

4. Keep classifications explainable

A useful classification records the triggering pattern, thresholds, affected interval, data-quality conditions, confidence, and alternative explanations. Terms such as probable outage or possible soiling communicate evidence without claiming a field-confirmed diagnosis. The recommended action should match the confidence: inspect communications first when data validity is weak, and inspect equipment when independent signals support a physical issue.

5. Use a repeatable triage order

  1. Validate timestamps, cadence, duplicates, and missing intervals.
  2. Check measurement ranges and frozen-signal behavior.
  3. Confirm that expected production is meaningful for the interval.
  4. Compare site, component, weather, and status evidence.
  5. Assign a probable category with confidence and a next inspection action.
  6. Quantify loss only across valid, comparable intervals.

Put the framework into practice

See how ArraySignal turns performance evidence into prioritized action.

Request a walkthrough using clearly labeled reference data and a workflow tailored to your portfolio context.

Book a Demo

Related resources

Continue the technical workflow.

Performance EngineeringStarter guide

A practical framework for expected vs actual PV performance

A field-oriented framework for comparing observed solar production with a defensible expected-production baseline.

8 min read
Read guide
Portfolio ManagementStarter guide

Building an energy-impact action queue for a solar portfolio

A practical method for ranking commercial PV sites by data confidence, potential energy loss, and operational urgency.

8 min read
Read guide