The real cost of a transformer over its life

Total cost of ownership is the sum of six cost elements, each incurred at a different time and each estimated differently. The table below sets out how each is normally quantified.

A workable evaluation needs four inputs before any bid is opened.

The first is the economic life. Use the life in the project model — typically 25 or 30 years for a utility-scale solar or wind asset — and apply it consistently to every cost stream. The second is the discount rate: the owner's real weighted average cost of capital, applied to real cash flows, with nominal rates and nominal prices kept on the other side of the calculation. The third is the duty cycle: annual operating hours, the equivalent load factor for loss purposes, and the harmonic spectrum at the point of connection. The fourth is the cost of an outage, which is the value of the energy not delivered while the transformer is out of service.

With those four inputs the model is arithmetic. The difficulty is not the arithmetic; it is that most tenders collect only the first two and then compare offers on purchase price.

U4E/UNEP, the United for Efficiency programme hosted by UNEP, states that transformers commonly stay in service for 30 years or more, and that the cost of the energy lost in them over that period can reach four times the purchase price. Two properties of the asset explain the ratio. A transformer is energised continuously, so no-load loss is paid for at night, at weekends and in every hour when the plant is idle. And it is a long-lived asset, so the loss stream is discounted over a very long period while the purchase price is paid once.

The practical consequence is that a specification which evaluates only the purchase price is optimising a minority of the cost. It also explains why loss capitalisation, the discounted valuation of the loss term, is usually the single sub-item of an LCC model that decides the award.

CIGRE Technical Brochure 939, published in 2024, analyses 425,000 transformer-years of service data from 66 utilities in 27 countries and 1,204 major failures. Its central finding for asset owners is that retirement risk accelerates after 20 years in service.

For a renewable project that finding sets a hard boundary. If the plant life is 25 to 30 years and the transformer is designed with the margins of a unit intended to be replaced at year 18, the owner will be buying a second transformer inside the offtake period. That replacement is not a normal capital event. Wood Mackenzie's Q2 2025 data puts power transformer lead times at around 128 weeks and generator step-up transformer lead times at around 144 weeks, so a failure at year 18 can mean close to three years without the asset.

The risk is not only thermal. It accumulates through moisture and ageing of the insulation system, through mechanical duty under short-circuit events, and through the corrosion and pollution conditions at the site. All of these are design inputs, and all of them are cheaper to address before the order than after.

Reliability enters an LCC model as an expected value: probability of failure in the year, multiplied by the cost of the consequences. CIGRE TB 939 provides the population failure rate; the bidder provides the data that shifts it — the thermal margin under the actual duty cycle, the dielectric margin, the short-circuit withstand evidence, the cooling redundancy, the condition-monitoring provisions and the guaranteed losses.

Three adjustments are worth making explicit. First, the design margin: a unit specified for the site's real ambient, altitude, seismic and pollution conditions has a lower expected failure rate than one specified for the defaults. Second, the spare strategy: a spare transformer is a capital cost that removes most of the outage term, and the two options can be compared directly. Third, the maintenance schedule: a design that requires less frequent oil treatment or tap-changer service is worth more in present value, and the bidder should be required to state the schedule rather than describe it qualitatively.

Where a bidder cannot supply the data, that is itself information. An offer that cannot state its expected maintenance interval or its cooling auxiliary power is an offer whose LCC cannot be evaluated.

Run at bid evaluation, an LCC model selects the cheaper of the designs on the table. Run earlier, it selects the design. That is the shift behind Lifetime-Matched Engineering™.

In practice the sequence is: the project's economic life, duty cycle and site conditions are fixed first; the transformer's thermal, dielectric and mechanical duty over that life is then modelled, using the thermal model of IEC 60076-7:2018 for oil-immersed units and the site's harmonic spectrum where it is significant; and the design is specified so that the margins are still adequate at the end of the plant's life rather than only at commissioning. On a solar or wind project that typically means a lower flux density and more core steel to hold no-load loss down over 25 years, more conductor cross-section to reduce load loss, a temperature-rise design that keeps the hot spot inside limits through the real duty cycle rather than at rated load, and cooling and protection selected for the conditions at year 20.

The verification side matters as much as the calculation. Losses, temperature rise and partial discharge are measured, and the measurements are what the owner will live with. Harbin Transformer has manufactured transformers since 1955 and operates a CNAS-accredited laboratory, registration L5393. Impulse testing is carried out with 1200 kV generators for products up to 220 kV and 4800 kV generators above that. The partial-discharge measurement background is 10 pC or lower on the oil-immersed line in Harbin and 1 pC or lower on the resin-insulated dry-type line in Hai'an, against the 100 pC limit that IEC 60076-3 sets for Um above 72.5 kV. A low measurement background is what makes a low partial-discharge guarantee verifiable.