Anatomy of a grid code test procedure: the nine sections regulators expect.

Conducting a grid code compliance campaign is a rare opportunity to get into the technical details of how a plant actually behaves — voltage response, frequency response, reactive capability, ride-through, protection coordination. But that opportunity only pays off if the campaign is prepared correctly. And preparation is a document: the test procedure.

Time invested in a strong procedure saves months of time and significant money later in the project. It reduces regulator query traffic. It reduces the risk of re-tests. It reduces contractual disputes between the plant, the EPC and the OEM. It reduces the risk of missing commercial operation.

At first glance, the requirement looks straightforward: comply with the applicable performance testing codes — ASME PTC 46 for combined-cycle capacity, IEC 61000-4-30 Class A for power quality, IEEE 2800 for inverter-based resources, and the grid code technical annex of the jurisdiction (in Mexico, Anexo 5 under the Manual de Interconexión de Centrales Eléctricas established by CRE Resolution RES/550/2021). But these codes give you a solid background — the general framework, the accepted instrumentation classes, the acceptable calculation methods. They cannot tell you how, when, and with what specific equipment to test your plant. That is what the site-specific procedure exists to do.

What a site-specific procedure is bridging

A procedure sits at the intersection of four constraints — and its job is to reconcile them into a single executable document:

• International testing codes — the technical foundation. ASME PTC 46 for combined-cycle acceptance, PTC 22 for gas turbines, PTC 6 for steam turbines. IEC 61000-4-30 Class A for power quality. IEEE 2800 for inverter-based resources. NIST-traceable calibration standards for instrumentation.
• Grid code technical requirements — the regulator's mandate. In Mexico, Anexo 5 of the Manual de Interconexión under CRE Resolution RES/550/2021. Equivalent frameworks in Brazil (Procedimentos de Rede via ONS), Chile (Norma Técnica de Seguridad y Calidad de Servicio via CEN), Colombia (Código de Conexión via XM), Peru (Norma Técnica de Calidad via COES), Morocco (Cahier des Charges via ANRE), Saudi Arabia (Grid Code via WERA).
• Contractual specifications — the guarantees the plant is being tested against. EPC contract guarantees. OEM performance guarantees. Lender-required demonstrations. Owner-required acceptance thresholds.
• Actual site conditions — the physics of your specific plant. Ambient temperature range, altitude, fuel LHV, water temperature, grid interconnection layout, instrumentation access points, communication paths, safety constraints.

Testing codes cover the first constraint in depth. Contract documents cover the third. But the second and fourth — the specific regulator requirements and the specific site conditions — are where every plant is different, and where the procedure has to do the reconciling engineering judgment that no external code can supply.

The nine sections a defensible procedure must contain

Across engagements on more than 50 plants — thermal, combined-cycle, hydro, renewable and BESS — the same nine sections show up in every procedure that has passed regulator review without a rework loop. When one of these sections is missing or thin, that is where the query traffic later concentrates.

1. Tests to execute — the enumerated list

The procedure begins with an unambiguous list of the tests to be performed on this plant. Not a paragraph. A table. One row per test, cross-referencing the applicable grid code section, the equipment under test (this unit, this transformer, this point of interconnection), the operating point (100% load, 75% load, at nominal voltage, at −5% voltage), and the acceptance criterion expressed as a number.

For a Type D combined-cycle plant under Mexico's Anexo 5, that list runs 25 to 45 individual tests per unit plus power quality tests at the interconnection point. For a BESS under IEEE 2800, it runs 18 to 30 tests. For a hydro plant, different. The enumeration is jurisdiction-specific and technology-specific — but it must be explicit. A procedure that says "grid code tests will be executed" is not a procedure; a procedure that says "Test 26 (Primary Frequency Response) will be executed on Unit 1 at 90% load with initial droop setting 4.0%, target response within ±0.5% of setpoint" is.

2. Instrumentation — accuracy class, model, calibration, location

Depending on the type of test, a specific level of test uncertainty is required. That uncertainty is driven almost entirely by the precision of the instrumentation. The procedure names each measurement channel and specifies:

• The accuracy class required by the applicable code (IEC 61000-4-30 Class A for power quality, ASME PTC 19.5 for flow, PTC 19.3 for temperature)
• The specific instrument model number that will be used and its current calibration certificate reference
• The exact installation location on the plant — with a P&ID or single-line reference
• The connection method — CT ratio, VT ratio, secondary wiring path, sampling rate
• The calibration traceability path (NIST for US, CENAM for Mexico, or the applicable national metrology institute)

A procedure that names instruments and cites their calibration is real. A procedure that says "calibrated equipment will be used" is not defensible.

3. Reference conditions — the baseline the measurement is corrected to

Grid code parameters fluctuate with weather and operating conditions. Ambient temperature affects gas turbine output. Barometric pressure affects steam turbine heat rate. Grid frequency affects governor response measurement. Fuel LHV affects heat rate calculation. The procedure specifies a set of reference conditions — typically ISO ambient (15 °C, sea level, 60% RH) for thermal capacity work, and nominal voltage/frequency for electrical performance work — to which measured values will be corrected.

Reference conditions are how the plant can compare tests over time and how the regulator can compare this plant to others. Without them, no meaningful comparison is possible.

4. Stability windows — how steady the plant must be for a valid data point

Every code specifies a stability window before a test point becomes valid data. ASME PTC 46 requires load, ambient conditions and fuel input to hold within specified tolerances for a minimum period before data collection begins. Grid code frequency and voltage tests have their own stability requirements — typically expressed as a maximum permitted rate of change of the parameter under test during the measurement window.

The procedure specifies, for each test, the stability tolerance and the minimum stable-run duration before data starts counting. This section is what protects the campaign from spending a week collecting data that later proves invalid.

5. Calculation methods — the formulas, cited to the code

The procedure includes a detailed section listing every calculation to be performed on the measured data, cited to the source code. Heat rate calculation cited to PTC 46. Uncertainty calculation cited to PTC 19.1. Droop calculation cited to the applicable grid code annex. Reactive capability curve cited to IEEE 421.5. Power quality aggregation cited to IEC 61000-4-30.

Where the grid code has its own specific formula (Mexico's Anexo 5 includes several bespoke ones), the procedure cites the grid code section directly. The regulator reads this section carefully — a formula in the procedure that cannot be traced to a code reference is the fastest path to a query.

6. Correction curves — how measured values are brought back to reference

Because measurement conditions vary and reference conditions are fixed, every test requires a set of correction curves. Ambient temperature correction for capacity. Altitude correction for combustion turbines. Fuel LHV correction for heat rate. Grid frequency correction for governor response. Cooling water temperature correction for steam turbine heat rate.

The procedure includes the correction curves themselves — usually as polynomial equations plus a graph — with the source of each curve identified. OEM-provided curves are cited to the OEM's performance document. Code-provided curves are cited to the applicable PTC or grid code section. Without complete correction curves, a measured value cannot be compared to a guarantee or to a threshold.

7. Applicable methods and codes — the full standards register

Individual activities within a testing campaign are governed by individual codes. Uncertainty calculation follows ASME PTC 19.1. Instrument calibration traceability follows NIST or the applicable national metrology institute. Data acquisition timing follows IEC 61869 for CTs/VTs. Fuel analysis follows ASTM D3588 or equivalent.

The procedure includes a table — the standards register — listing every code applied, its edition/year, and the campaign activities it governs. This section is what makes the dossier auditable years later, when a lender or an acquirer does technical due diligence and needs to confirm the campaign followed accepted practice.

8. Data collection protocol — sampling, aggregation, storage

During the test, data collection must follow a specified protocol. Not general guidance — a specification. For each data source, the procedure names:

• The sampling rate (typically 200 ms base for grid code work, aggregated to 10 min and 2 hr per IEC 61000-4-30)
• The timestamping standard (UTC, PTP-synchronized where required)
• The storage format (CSV structure, PQDIF file, PMU stream)
• The channel list with tag names
• The backup and integrity protocol (parallel logging, checksum verification, chain of custody)

A procedure that specifies the file format, the time base, the sampling rate, the channel list and the storage protocol is dossier-ready. A procedure that says "data will be recorded" is not.

9. Report templates — the deliverable the regulator will actually read

The procedure includes the templates for the reports that will be produced from the campaign. Not descriptions of the reports — the actual templates, with section structure, expected data tables, expected plots, cover page, sign-off blocks and revision history.

The regulator reads reports, not raw data. Reports that arrive in the regulator's expected format — SAPPSE structure for Mexican CENACE submissions, ONS-Web structure for Brazil, XM's platform for Colombia, ANRE's format for Morocco — clear review in one or two revisions. Reports that arrive in a novel format spend six months in query traffic while the regulator asks for reformats. The templates are pre-negotiated by including them in the procedure and having the procedure approved.

What happens when a section is missing or thin

Each of the nine sections exists because a specific class of failure occurs when it is missing. The failure modes are predictable and, once seen, obvious in hindsight.

• Missing enumerated test list — scope disputes with the regulator mid-campaign. "You said grid code testing but you did not specify this test." Program stops until the scope is renegotiated.
• Missing instrumentation specification — data rejected during dossier review. "We cannot accept this measurement because you have not demonstrated the accuracy class." Re-test at owner's cost.
• Missing reference conditions — measured values are not comparable to guarantees. Contractual dispute between plant, EPC and OEM about whether the guarantee was met.
• Missing stability tolerances — a week's worth of data is invalidated because the plant was not stable when it was collected. Re-test.
• Missing calculation formulas — regulator queries every calculated value. Query traffic explodes. Program duration extends by weeks.
• Missing correction curves — measured values cannot be brought to reference. Every test result becomes negotiable rather than definitive.
• Missing codes register — dossier fails a due-diligence audit later, when a lender or acquirer looks at whether the campaign followed accepted practice.
• Missing data collection protocol — raw data is captured in a format that cannot be reprocessed later if a question comes up.
• Missing report templates — reports arrive in the wrong format for the regulator and spend months in reformatting cycles.

Each of these failure modes is preventable. The engineering investment to prevent them is orders of magnitude smaller than the cost of experiencing them.

The compounding effect

None of the nine sections is difficult to write on its own. The difficulty is that they must be internally consistent — the instrumentation section must match the calculation section, which must match the correction curves, which must match the report templates. And all of them together must match the applicable code, the contract, and the actual plant.

This is why a good procedure takes six to twelve weeks of experienced engineering time to write. Not because any single section is complex, but because getting them consistent — and then keeping them consistent through two or three review cycles with the regulator — requires deliberate work.

The engineering investment shows up on the other side. Plants whose procedures are strong reach commercial operation with clean dossiers, minimal query traffic, no re-tests and no scope disputes. Plants whose procedures are weak spend the equivalent of the procedure-authoring cost many times over in delayed COD, extended defense, contractual disputes and lender friction. The procedure is where the leverage lives.

The nine sections of a procedure are the first artifact of any grid code testing engagement. If yours is being authored in-house or by an EPC, an independent review of the document — before it reaches the regulator — is usually the highest-leverage engineering hour a plant owner spends during commissioning.