PRC-005 Relay Maintenance Evidence: Where AI Review Helps and Stops

NERC PRC-005-6 is often shortened in conversation to “the relay testing requirement,” but that shorthand can blur important distinctions. The standard concerns a maintenance programme for protection systems, automatic reclosing, and sudden-pressure relaying for entities and facilities within its scope. A programme includes scope, intervals, activities, procedures, qualified personnel, records, and corrective action. A software product does not become a compliance certificate because it can export a PDF.

What an evidence workflow should preserve

For each maintenance activity, the record should make the following visible:

• asset and scheme identity;
• applicable maintenance activity and interval;
• relay, communication, auxiliary, and trip-circuit scope;
• procedure version and test method;
• test technician qualification where the procedure requires it;
• instrument identity and calibration evidence;
• settings and configuration version;
• expected, measured, tolerance, and verdict data;
• as-found condition, corrective action, and as-left condition;
• exceptions, missed dates, re-tests, and approval.

The exact fields belong to the utility’s programme and the applicable standard. The principle is general: a later reviewer must be able to tell what was maintained and how the result was accepted.

Common evidence gaps

In practice, teams often have the data but not the relationship. A test result sits in a PDF folder, the calibration record is in a spreadsheet, the settings are in a relay-management tool, and the work order is in a CMMS. The individual files may be correct; the audit trail is still difficult to assemble.

Other gaps include:

• a test date without a clear interval or due date;
• an asset name that changed after a station reconfiguration;
• a “pass” with no tolerance or procedure version;
• a new settings group tested without retaining the as-found export;
• a failed point corrected on site with no record of the original failure;
• a communications or trip-path test noted in prose but not linked to raw evidence;
• a report signed by a person whose role or approval is not recorded.

The GridAPM PRC-005 evidence checklist is useful for organizing questions, but it should not be treated as a universal compliance template.

What AI review can do

A bounded assistant can scan a maintenance evidence pack and return a review queue:

1. identity missing or inconsistent;
2. settings version not linked to the test;
3. interval cannot be calculated from the available dates;
4. calibration status absent or outside the procedure’s requirement;
5. test case listed but no result file found;
6. failure or deviation with no disposition;
7. as-found/as-left status unclear;
8. report references a standard or procedure without a source.

It can also assemble a draft package from linked records and explain where a value came from. The control is that it must show the source, confidence or uncertainty state, and the exact gap. It must not label a missing record compliant or invent a date.

Maintenance is not only relay injection

Protection maintenance may include the relay, instrument transformers, DC system, trip circuits, communications, breaker interface, automatic reclosing, and sudden-pressure functions depending on the programme. The Bureau of Reclamation’s FIST material emphasizes field procedures, qualified personnel, peer review, site knowledge, and the need to verify the full path where the procedure calls for it.

An AI system that sees only the relay’s internal test result cannot claim to understand the complete protection system. The evidence model should mark component coverage and identify what was not tested. A gap is not a failure; hiding the gap is the failure of the record.

PRC-004 and the feedback loop

Maintenance evidence and misoperation analysis should inform each other. PRC-004 addresses identification and correction of protection-system misoperations. A recurring test deviation, a settings mismatch, or an event record may reveal a common cause across multiple assets. The maintenance work package should carry the relationship between the original evidence, engineering assessment, corrective action, and re-test.

An agent can help find that relationship across records. It cannot determine the cause without an engineer testing the hypothesis against the system, settings, device, and procedure.

OT security boundary

If maintenance evidence is processed with a hosted AI provider, the organization needs a data-flow decision: what context leaves the OT boundary, what identifiers are removed or retained, who controls the provider key, and how the result is recorded. NIST SP 800-82 is a useful reference for the difference between ordinary IT processing and systems that monitor or directly affect the physical environment.

The safest default is a read-only workbench outside the protection path, with explicit outbound context for each requested AI operation and no ability to write relay settings or control equipment. “Local-first” is a design claim that still needs architecture evidence and customer review.

ProtectionAI’s honest role

ProtectionAI can support the clerical and analytical layer around relay maintenance: organizing approved evidence, finding inconsistencies, drafting a report, and preserving reviewer context. It should be presented as a support tool for a customer-controlled maintenance programme. The product does not certify PRC-005, replace required physical tests, or remove the need for qualified engineers.

For a pilot, measure whether the tool reduces missing links between work orders, settings, raw files, and signed reports. Sample the results manually and record false flags as well as useful findings. The quality metric is not how many documents the AI touched; it is whether the final evidence is clearer, more complete, and easier for the responsible engineer to defend.

References

• North American Electric Reliability Corporation. (2025). Protection system, automatic reclosing, and sudden pressure relaying maintenance (PRC-005-6). https://prod.nerc.com/standards/reliability-standards/prc/prc-005-6
• North American Electric Reliability Corporation. (n.d.). Protection system misoperation identification and correction (PRC-004-6). https://www.nerc.com/globalassets/standards/reliability-standards/prc/prc-004-6.pdf
• Kezunovic, M. (2021). Life cycle testing of synchrophasor based systems used for protection, monitoring and control (CIGRE Technical Brochure No. 843). https://www.e-cigre.org/publications/detail/843-life-cycle-testing-of-synchrophasor-based-systems-used-for-protection-monitoring-and-control.html
• U.S. Department of the Interior, Bureau of Reclamation. (2011). Operation, maintenance, and field test procedures for protective relays and associated circuits (FIST 3-8). https://usbr.gov/power/data/fist/fist3_8/vol3-8.pdf
• Stouffer, K., et al. (2023). Guide to operational technology (OT) security (NIST SP 800-82 Rev. 3). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.SP.800-82r3