The Second Specification: From Site Conditions to Acceptance

The first specification is a purchasing document, and it is good at what it does. It fixes the voltage class and the highest voltage for equipment (Um), the rated power, the vector group and impedance, the tapping range, the applicable standard — IEC 60076-1 in most markets, AS/NZS 60076.1:2014 in Australia and New Zealand, the GB/T 1094 series in China — the loss limits where a market regulates them (IEC TS 60076-20), and the test regime — the document a manufacturer needs in order to quote.

What it usually leaves out is everything the transformer will experience after it is energised, and everything the buyer will need in order to accept it. Ambient temperature is replaced by the IEC 60076-2 reference conditions — maximum 40 °C, monthly average 30 °C, annual weighted average 20 °C — even when the site records none of those numbers. The harmonic environment is left to the inverter supplier's data sheet. The transport route, the documents required at handover, the spare parts list and the training scope are usually absent.

The most common failure is a specification carried over from a previous project. A document written for a European reference climate and reused at a site with 2,400 m altitude, 50 °C ambient or 0.4 mg/cm² equivalent salt deposit density specifies the wrong equipment with the right voltage. Insulation levels are corrected for altitude above 1,000 m under IEC 60071-2, and pollution performance is selected under IEC 60815-1 and IEC 60815-2 rather than assumed.

The second failure is a nameplate specification with no duty cycle behind it. A storage transformer that cycles two or three times a day, or a wind turbine transformer at partial load for most of its hours, has a thermal duty the rated power does not describe. The third is harmonics. Inverter and PCS switching produces a current spectrum above fundamental frequency, which raises eddy and stray losses and can overheat a winding that is still inside its nameplate rating. IEEE C57.110 exists to establish capability under non-sinusoidal load current, and IEEE 519-2022 defines the limits at the point of common coupling; neither can be applied if the specification never states the converter model or the expected spectrum.

The second specification is a site-and-acceptance document. It should state the items below — and for each one the source of the information, because a condition without a source is an assumption, not a requirement.

A condition only becomes enforceable when it is attached to a design margin and a test. Each line of the second specification should therefore name the clause it is designed to and the evidence that will demonstrate compliance.

The last row needs a note. Short-circuit withstand testing under IEC 60076-5 is destructive, is performed on a sample basis, and takes place at an external high-power laboratory. What a buyer receives is therefore a type-test report for the design family, not a routine test on their own unit. Read it for the tested rating, the tested impedance and the test duration, and confirm that the tested configuration matches what is purchased. The same discipline applies to partial discharge: IEC 60270 defines the measurement, with a measuring band of 30 kHz to 1 MHz, and IEC 60076-3 sets the limit — 100 pC for units with Um above 72.5 kV, although many owners now write 50 pC. The specification should state both the limit and the background noise level at which it will be measured.

Losses belong in the specification as an evaluation rule, not as an aspiration. State the capitalisation factors — a currency value per kW of no-load loss and per kW of load loss — in the tender documents, require bidders to declare no-load and load loss at reference temperature to IEC 60076-1, and evaluate on price plus capitalised losses. A manufacturer that knows how losses will be weighted will design for the weighted cost; one that does not will design for the cheapest unit that passes. The method is set out in our note on loss capitalisation.

The opposite error is over-specification. Naming a specific bushing or tap changer brand, demanding an impedance outside the useful range, or requiring a loss figure that only one design can reach narrows the bidder pool. In a market where power transformer demand exceeds supply by roughly 30%, narrowing the pool has a price, usually paid in the delivery date rather than the bid. Specify performance, interfaces, dimensions and evidence; leave the internal design to the manufacturer.

The second specification is not an appendix to the purchase order. It decides whether the transformer is still the right unit in year 20, which is why we treat it as design input, not documentation.

In practice, the site conditions are read first and the electrical specification is read against them: each condition is traced to a design margin and a test, and the acceptance list is agreed before the order is placed, so the test programme is planned rather than negotiated at the end of manufacture. Those tests — impulse at the corrected insulation level, temperature rise at the site ambient, partial discharge at the specified limit — are the ones that consume test-bay time. Our two impulse generator systems, 1200 kV for products up to 220 kV and 4800 kV above that, and partial discharge backgrounds of ≤10 pC for oil-immersed units at Harbin and ≤1 pC for dry-type units at Hai'an, are the evidence that those acceptance items can be executed rather than promised; the laboratory holds CNAS accreditation L5393.

CIGRE TB 939 (2024), drawing on 425,000 transformer-years, 66 utilities and 1,204 major failures, found that retirement risk accelerates after 20 years in service; U4E/UNEP notes that operating losses can reach four times the purchase price over a service life of 30 years or more. A specification that stops at the nameplate has already given up on both. Built for year 20, not commissioning day.