1. What a performance test actually is
A performance test is a controlled measurement of a power plant's two commercial numbers — output (how much power it delivers) and heat rate (how much fuel it burns to deliver it) — carried out under defined conditions and compared against the guarantees in the contract. Everything else is commentary. When a new plant is handed over, the test decides whether the EPC contractor met the guarantee or owes liquidated damages. When an operating plant is tested, the result feeds tariffs, fuel procurement, degradation tracking and overhaul justification.
A test is not a snapshot of the plant's SCADA trends on a good day. It is a formal exercise with an agreed procedure, a defined measurement boundary, calibrated instrumentation, stability criteria, and results corrected to the reference conditions written into the guarantee. The difference between a defensible test and an expensive argument is almost always decided before the test window opens — in the procedure, not on the day. My article on what a performance test really proves walks through that distinction in plain words.
2. The standards landscape
| Code | Scope | Used for |
|---|---|---|
| ASME PTC 46 | Overall plant performance | Whole-station output and heat rate — the usual acceptance basis for combined cycle and thermal plants |
| ASME PTC 22 | Gas turbines | GT output and heat rate, standalone or as part of a plant test, including firing-temperature-limited machines |
| ASME PTC 6 | Steam turbines | ST output and steam rate with the tightest uncertainty discipline in the family |
| ASME PTC 4 | Fired steam generators | Boiler and HRSG efficiency, input-output and heat-loss methods |
| ISO 2314 | Gas turbines | The ISO counterpart to PTC 22, common in European-supplied contracts |
The contract names the code, and the code governs everything downstream: which quantities are measured, how corrections are derived, and how uncertainty is treated. Two further documents matter as much as the code itself — the guarantee schedule (the numbers and the reference conditions they apply at) and the OEM correction curves (the manufacturer's own statement of how the machine responds to ambient temperature, pressure, fuel properties and load). A test planned without all three documents on the table is not yet a test.
3. Preparation: where tests are won and lost
The procedure
The written test procedure translates the code into a plant-specific script: test boundary diagram, instrument list with ranges and calibration status, operating configuration (which auxiliaries run, which valves are lined up), the load points to be held, the duration of each run, and the stability criteria that make a run valid. All parties — owner, contractor, and any independent witness — sign it before the window. Changes after that are variations, and they get documented as such.
Instrumentation and the boundary
Test accuracy is bought with temporary instrumentation: calibrated pressure and temperature elements at the boundary points, a revenue-grade measurement of electrical output, and a defensible fuel measurement — flow, composition and heating value sampled during the run, not taken from a monthly average. The boundary defines what the numbers mean: gross output at the generator terminals and net output at the grid export point are different products, and heat rate follows the same boundary. The classic disputes — auxiliary loads inside or outside, fuel measured at which flange — are boundary disputes, settled by the diagram, not by debate after the fact.
Stability and the run itself
A valid run requires the plant to be demonstrably steady: output, temperatures and pressures inside agreed bands for the run duration, no operator interventions that the procedure doesn't allow, and the configuration frozen as specified. Data is typically averaged over one to two hours per load point. A run that fails its stability criteria is discarded and repeated — an uncomfortable rule that protects both sides.
4. Corrections to reference conditions
Guarantees are written against reference conditions — often ISO conditions (15 °C, 101.325 kPa, 60% RH) for gas turbines, or contract-specific values for the whole plant. Tests happen on real days. A summer test in Karachi at 41 °C and sea-level pressure is not the reference day, and neither party should be judged by the weather. The correction method multiplies each measured quantity by factors derived from the OEM correction curves:
Output on a hot day is corrected upward; heat rate on a hot day is corrected downward, because the machine was fighting the ambient. Four disciplines keep corrections honest: the curves must be the contract curves for that machine, factors are computed for the exact test-point conditions, corrections for output and heat rate are kept consistent with each other, and estimated factors are never reported as test results. The mechanics, with a worked example, are in the Performance Correction Calculator.
5. Uncertainty — the number inside the number
Every test result carries an uncertainty band built from instrument errors, sampling variation and correction-curve tolerance. PTC methods quantify it; what happens next is contractual. Some contracts compare the corrected result directly against the guarantee. Others apply a tolerance, or require the result minus uncertainty to clear the line. This single clause can move millions, because a plant that "misses by 0.4%" may be inside its own measurement uncertainty. Professional practice on both sides: agree the treatment of uncertainty in the procedure, quantify it from the actual instruments used, and never discover the clause after the test.
6. Heat rate, degradation and the tariff connection
After acceptance, the same numbers keep working. The tested heat rate typically becomes the reference for the plant's tariff — in Pakistan, fuel cost components in the tariff are built on demonstrated efficiency, as explored in From Heat Rate to Tariff. Repeat testing then separates the three causes of declining performance: recoverable fouling (fixed by washing and cleaning), repairable degradation (fixed by overhaul), and the ambient and loading conditions the plant cannot control. The Degradation Tracker and the Heat Rate & Efficiency Converter are the working tools for that follow-up, and fuel-side calculations — flow from output and heat rate, gas heating value and Wobbe index, orifice metering — sit in the same toolkit: Fuel Flow, Gas Properties, Orifice Flow.
7. How tests go wrong
- Boundary creep — auxiliary loads or fuel take-offs quietly crossing the test boundary between procedure and run.
- Fuel data by assumption — heating value taken from a certificate or a monthly average instead of samples drawn during the run.
- Correction curves from the wrong machine — a sister unit's curves, or a newer revision than the contract set.
- Stability theatre — holding impressive averages while the plant drifts inside them; the criteria exist to catch exactly this.
- Uncertainty discovered late — the tolerance argument started after the result is known.
Every one of these is a preparation failure, which is the optimistic reading: all of them are preventable in the procedure stage.
8. Frequently asked questions
What is the difference between PTC 46 and PTC 22?
PTC 46 covers overall plant performance — the whole station's output and heat rate. PTC 22 covers the gas turbine alone. A combined cycle acceptance test typically runs under PTC 46 for the plant guarantees, with PTC 22 methods applied to the gas turbines within it.
How long does a performance test take?
The runs themselves are measured in hours per load point, but a properly prepared test is a two-to-six-week exercise end to end: procedure agreement, instrument installation and calibration, the test window, then data reduction, corrections and the report.
Can a failed test be repeated?
Contracts usually allow a retest under defined conditions, sometimes after corrective work by the contractor. What cannot be done is re-running until the numbers please someone — each attempt is documented, and stability and procedure compliance decide validity.
Who should witness a performance test?
Both contracting parties, and wherever the result goes to a regulator, lender or offtaker, an independent test agency or owner's engineer whose only interest is that the procedure was followed. Independence is what makes the report usable by everyone.
9. Work with me
I lead performance tests and defend their results in front of regulators — most recently at 900 MW and 1,263 MW scale, with reports submitted to NEPRA and CPPA-G. If you are planning an acceptance test, disputing one, or setting up a repeat-testing regime, the services page sets out how I take on that work, or you can go straight to the contact page.
Saad Hafeez is a Chartered Mechanical Engineer (CEng MIMechE), Deputy General Manager – Power Plants at OMS, with 15+ years across combined cycle, gas turbine and reciprocating engine plants.
Engineering calculators | Guide: Power Plant Maintenance | Services