Oil & Gas Refinery Transformer: How to Size the Substation for a Process Unit, Why Hazardous Area Classification and Motor Starting Decide the Spec, and When Dry-Type Beats Oil

The answer first: a refinery or petrochemical transformer is sized from the process load list and from the consequence of losing a process unit, not from a kVA-per-barrel rule of thumb. What actually decides the rating is a combination of three things that have nothing to do with average demand: the largest compressor or pump motor that has to start while the rest of the plant is running, the amount of non-sinusoidal current that variable-frequency drives push back through the transformer, and the fact that the transformer usually has to sit outside the hazardous area while the load it feeds sits inside it. A competent specification therefore starts with a load list by process unit, applies demand and diversity, checks the voltage dip at the largest motor start, derates for harmonic loss, and only then selects a standard rating and an impedance that keeps the switchgear and the arc-flash energy inside their limits.

A refinery is not a single load. It is a chain of process units — crude distillation, vacuum distillation, hydrotreating, reforming, fluid catalytic cracking, delayed coking, sulphur recovery, product treating, blending, and the utilities that hold the whole site together — and each unit has its own mix of continuously running pumps, large compressors, fired heaters, air coolers and instrumentation. The electrical system that serves them is built in tiers, from the utility intertie down to 0.4 kV motor control centres, and a transformer sits at every boundary between those tiers. Sizing the wrong tier is how a project ends up with a 40 MVA transformer that cannot start its own compressor, or with an 11 kV switchboard that cannot be safely operated because the upstream impedance was chosen for loss efficiency instead of fault level.

This guide is written for the people who actually have to specify and buy that equipment: a refinery electrical engineer preparing a datasheet, an EPC contractor building the single-line diagram, or a procurement team comparing offers from three manufacturers who have quoted three different impedance values and three different cooling classes for the same duty. It follows the same method we use when sizing equipment for cement, mining and desalination plants, applied to the specific constraints of hydrocarbon processing: explosive atmospheres, very large motors, high harmonic content, and a site that cannot afford an unplanned shutdown. If you are sizing a generic industrial distribution system instead, start with our distribution system sizing guide and come back here for the hazardous-area and motor-starting detail.

Key takeaways

• Size from the load list, not from throughput. Build the list unit by unit, count spared pumps once, add 10–20% design margin, and size the transformer for the worst credible motor-starting case rather than for the arithmetic sum of nameplates.
• Hazardous area classification moves the transformer, it does not usually change its certificate. Oil-immersed power transformers are rarely Ex-certified; the usual solution is a non-hazardous or pressurized substation with a fire wall and a safe-air intake, designed under IEC 60079-13 and API RP 540.
• The largest motor decides the transformer, not the total kVA. A 10 MW compressor started direct-on-line at 5.5× locked-rotor current can pull more reactive power than the rest of the substation combined, and will dip the bus by 15% or more unless the impedance, the starting method or the transformer rating is changed.
• VFD load is not nameplate load. A drive-heavy refinery bus needs either a K-rated or harmonic-derated transformer, or mitigation at the drive, because the eddy and stray losses rise with the square of the harmonic order and the winding hotspot rises faster than the average temperature.
• Impedance is a three-way trade. Higher impedance lowers fault current and arc-flash energy but worsens voltage dip, motor starting and regulation; 7.5–10% on a 25–40 MVA, 33 kV unit is the usual compromise, and it must be checked against the switchgear rating and the largest motor, not picked from a catalogue.

What does a refinery transformer actually feed? The voltage architecture in five tiers

Almost every refinery and petrochemical site in the world uses the same five-tier structure, with slightly different numbers depending on whether the site was built to IEC or to North American practice. The transformer at each boundary has a different job, and confusing the jobs is the most common source of oversizing.

Two structural differences separate a refinery from a normal factory. First, the MV level is usually a motor bus, not a distribution bus: 11 kV or 6.6 kV feeders go directly to compressors, pumps and fans, so the transformer sees motor starting transients many times a shift. Second, the LV level is fragmented into dozens of small substations scattered through process areas, each one subject to a hazardous area classification that the transformer must respect even though the transformer itself is a passive, sealed piece of equipment.

How do you build the load list for a process unit without over- or under-sizing?

The load list is the only defensible basis for the rating. A refinery load list is built by unit, not by plant, and each unit is built by consumer type: continuously running pumps (often spared), large compressors (usually not spared), air coolers, fired-heater auxiliaries, instrument air, and utility loads. The method is mechanical once the discipline is in place:

• List every consumer with its rated kW, efficiency, power factor, duty (continuous, intermittent, standby) and starting method. For motors above roughly 1 MW, the locked-rotor current and starting power factor matter as much as the rated kW.
• Count spared pumps once. A duty/standby pair is one running load plus a standby that is started, not two running loads. Adding both is the single most common cause of a 30% oversized substation. The standby unit does matter for starting studies and for the N-1 check.
• Apply a demand factor per unit and a diversity factor across units. Typical refinery practice lands at 0.85–0.95 demand within a unit and 0.85–0.95 diversity across units, because units rarely peak simultaneously; the loss of the largest single consumer should not be assumed, since a trip is exactly the case you are designing for.
• Add design margin and spare capacity. 10–15% for load growth over the first five years, plus a deliberate spare feeder or two for future debottlenecking, is normal; 40% blanket spare is not, and it costs loss capitalisation for twenty years.
• Convert to kVA at the design power factor. Refineries run at 0.85–0.90 lagging before compensation, lower during motor starting; if power factor correction is installed, correct at the MCC or at the MV bus and do not assume the correction is in service during a start.

As a sanity check, not as a design basis: a 100,000 barrel-per-day refinery typically draws somewhere in the range of 60–100 MW, and a 200,000 bpd site with conversion units (FCC, coker, hydroprocessing) draws on the order of 150–250 MW. Those figures are industry experience values for order-of-magnitude checking only. They should never replace a load list, because two refineries of identical throughput can differ by 40% in electrical demand depending on configuration, hydrogen supply, and how much steam versus electric drive the design uses.

The load figures above are industry experience ranges for a mid-size refinery and vary widely with configuration, climate and driver choice. Use them to check the order of magnitude of a load list, never to replace one.

Why does hazardous area classification change the transformer specification more than the kVA does?

Because the transformer is almost never the piece of equipment that carries the Ex certificate. A power transformer is a sealed, oil-filled or cast-resin block with no arcing contacts in normal service, and Ex-certified power transformers at 10–40 MVA essentially do not exist as a catalogue product. What the classification actually controls is where the transformer may be placed, and therefore what separates it from the process, what fire protection it needs, and how its cables enter and leave.

Under IEC 60079-10-1 a release of flammable gas or vapour creates Zone 0 (continuously present), Zone 1 (likely in normal operation) or Zone 2 (unlikely, and only briefly). Under NEC Article 500 and API RP 500, the equivalent concepts are Class I Division 1 and Division 2; Article 505 maps the same physics onto Zones, and Articles 506 and the IEC 60079-31 dust series cover combustible dust. Refineries and petrochemical plants are, in practice, mostly Zone 2 / Division 2 territory, with Zone 1 / Division 1 confined to small envelopes around pump seals, sample points, compressor seals and open drains. That distinction matters commercially: a transformer placed 3 metres outside a Zone 2 boundary costs a fraction of one that has to be installed inside it.

Three details are routinely missed. First, the cooling air intake must come from an unclassified area, otherwise the pressurization concept is defeated the moment the wind changes; this frequently forces the substation to the upwind face of the unit. Second, cable penetrations need sealed glands and, for a pressurized room, a documented leakage rate so the pressurization system can hold the required overpressure. Third, the oil-filled transformer itself is a fire source, so the fire wall, the blast wall and the separation distance are part of the electrical design, not an afterthought for the civil contractor — API RP 540 and NFPA 30 both treat transformer oil as a combustible liquid inventory that must be contained and drained away from the process.

Oil-immersed, cast-resin dry-type, or a pressurized substation: which belongs inside a process unit?

The choice is driven by fire load and by ambient, not by efficiency alone. Oil-immersed units are more efficient, cheaper per MVA, and available in sizes up to and beyond 100 MVA, but they bring a combustible liquid inventory and a containment requirement. Cast-resin dry-type units remove the oil and the containment, tolerate moisture and dust better than older dry types, and can be installed closer to the load, but they are limited in practice to a few MVA per unit at 11–33 kV, cost more per MVA, and are less efficient at the same rating.

The decision rule that survives contact with real projects: put oil where the MVA is, put dry-type where the fire load is unacceptable, and put both behind a wall if the classification boundary is closer than the fire separation distance allows. For a deeper comparison of the electrical trade-offs, including impedance and short-circuit behaviour, see our oil-immersed vs dry-type selection guide.

Motor starting: can a 10 MW compressor start across the line on this transformer?

Usually not, and the arithmetic is simple enough to do before the purchase order. Take a 10 MW, 11 kV synchronous motor with 0.97 efficiency, 0.92 power factor and a locked-rotor current of 5.5 times rated. Its rated input is about 11.2 MVA and its full-load current about 588 A. At start it draws roughly 3,235 A at a starting power factor near 0.2–0.25, which is a starting apparent power of about 61.6 MVA — more than five times the rating of the transformer that is supposed to feed it.

Voltage dip follows directly from the source strength. If the upstream 25 MVA transformer has 8% impedance, its short-circuit capacity at the 11 kV bus is 25 / 0.08 = 312 MVA, and the dip during the start is approximately starting MVA divided by (starting MVA + source MVA), or 61.6 / (61.6 + 312) = 16.5%. That is beyond the 10% usually accepted at an MCC bus and close to the 15% that motor terminals will tolerate, and it is computed with every other motor already running. Add the running load, and a plant that assumed "10 MW on a 25 MVA transformer is 40% loaded, so it is fine" discovers at commissioning that the compressor trips the upstream breaker on undervoltage.

Dips calculated as starting MVA / (starting MVA + source short-circuit MVA), assuming the motor is the only load switching and the source is a single 25 MVA transformer at 8% impedance. Real studies add running load, motor contribution and cable impedance, and should be done in a load-flow and motor-starting program, not by hand.

The four ways out, in the order a refinery normally considers them:

There is a second-order effect that catches designers: the transformer that feeds a VFD is not sized by the motor's starting current but by its harmonic current and by the drive's own overload profile. A VFD eliminates the dip and creates a harmonic problem, and the harmonic problem is what the transformer datasheet has to reflect.

How much harmonic current do refinery VFDs actually produce, and how do you size for it?

Refineries are among the most drive-heavy industrial sites in the world, because pumps and compressors are the loads and speed control is where the energy savings are. A 6-pulse drive without mitigation draws a current waveform with roughly 25–40% total harmonic distortion, dominated by the 5th and 7th harmonics, with the 11th and 13th next. A 12-pulse drive halves that to roughly 10–15%, and an active front end or an active filter brings it to about 3–5%. Those are typical values from industry experience and vary with the drive's DC link, the source impedance and the load point.

IEEE 519-2022 limits the distortion the site injects at the point of common coupling, typically 5% total demand distortion for a strong bus, with individual harmonic limits that tighten as the short-circuit ratio rises. That standard protects the grid. It does not protect the transformer. For the transformer the governing standard is IEEE C57.110, which deals with the fact that eddy-current and stray losses rise with the square of the harmonic order: a 5th harmonic contributes 25 times the eddy loss of the fundamental for the same current, an 11th harmonic 121 times. The result is a winding hotspot that runs far hotter than the average temperature suggests, and a transformer that fails its temperature-rise test at what looked like 80% load.

There are two ways to specify the answer, and the correct one depends on who is buying:

• K-factor rating. Specify a K-rated transformer (K-4, K-9, K-13 or K-20) whose winding is designed for the eddy losses of that harmonic spectrum. K-4 suits a bus with a minority of 6-pulse drives; K-13 suits a bus that is mostly drives. K-rated units are usually dry-type and usually limited in MVA.
• Derating and loss specification. For large oil-immersed units, keep the standard design, compute the harmonic loss factor F_HL from the actual spectrum, and derate the permissible current per IEEE C57.110. In practice a 6-pulse-dominated bus without mitigation lands at a permissible current factor of roughly 0.8–0.9, which means 10–25% more kVA than the fundamental load list suggests. The alternative is mitigation at the drive: a 4% line reactor, a DC-link choke, a 12-pulse configuration or an active front end.

Do not double count. If the drives are specified with active front ends, the harmonic derating on the transformer largely disappears and the transformer can be sized on fundamental current plus a small margin. If the drives are plain 6-pulse, the derating is real and must appear on the datasheet as a specified harmonic spectrum, not as a vague note that "the load contains drives". For the underlying physics, see our transformer harmonics guide and, if the bus is dry-type, the K-factor selection guide.

What impedance should a refinery transformer have?

Impedance sets three things at once: the fault current the downstream switchgear must interrupt, the arc-flash energy a technician is exposed to, and the voltage dip and regulation the process sees. Raise it and the switchgear gets cheaper and safer; lower it and the motors start more easily and the voltage holds better. On a 33/11 kV, 25 MVA unit, 7.5–8% is a common refinery choice precisely because it puts the 11 kV bus in the 16–20 kA range, which matches standard 25 kA switchgear with margin.

Two cautions. First, impedance has a tolerance: IEC 60076-1 allows the measured value to deviate from the declared value, and the negative tolerance is what governs the fault study, so specify the tolerance and use the lower bound in the calculation. Second, a high-impedance transformer reduces arc-flash incident energy but does not eliminate it; IEEE 1584-2018 calculations still have to be done at every bus, and the result usually drives the choice between a 25 kA and a 40 kA switchboard more than any other single number. If you are working through the fault calculation itself, our short-circuit current guide walks through the three-step method and the impedance tolerance trap.

How much redundancy does a refinery transformer really need?

The answer is set by the consequence of losing the unit, and refineries usually answer it with one of three arrangements. The choice is a capital-cost decision that should be made explicitly, because the default in many FEED packages is 2×100% everywhere, which is both expensive and, at the LV level, frequently pointless.

The load factor matters as much as the redundancy. A refinery transformer that runs at 0.85–0.95 of rating for 8,000 hours a year ages faster than a distribution transformer that runs at 0.6, and the Arrhenius relationship means the loss of insulation life is not linear. If the load list says the unit will sit above 0.85 for most of the year, either size up by one standard step or specify a lower temperature rise so the hot spot stays inside the class. Our load factor and overload guide covers the thermal arithmetic.

Heat, corrosion and fire: what changes for a 50 °C coastal site?

A refinery in the Middle East, West Africa or coastal Southeast Asia combines three stressors that a European datasheet does not reflect: 45–50 °C ambient with intense solar radiation, salt-laden air at ISO 12944 C5-M corrosivity, and a process fire hazard. Each one moves a specification.

• Ambient. Standard ratings assume 40 °C maximum ambient. Above that, an ONAN unit loses roughly 1% of rating per degree Kelvin, so a 50 °C site derates an air-cooled unit by about 10% before any harmonic derating. Specifying a second ONAF stage, or a lower temperature rise (55 K instead of 60 K top-oil), buys back capacity at the cost of a larger radiator bank.
• Solar radiation. A transformer in direct sun sees an effective ambient several degrees above air temperature; shading, a sun shield or an indoor placement is worth more than an extra cooling stage in some designs.
• Corrosion. C5-M requires a documented paint system (typically a zinc-rich primer, an epoxy intermediate coat and a polyurethane topcoat, 240–320 µm total), stainless-steel or hot-dip-galvanised hardware, and sealed cable boxes. Radiator fins are the first thing to fail; specify their thickness and coating explicitly.
• Fire. Oil containment sized to 110% of the oil volume, a fire wall or blast wall to the process, drainage away from the unit, and a deluge or foam system where the separation distance cannot be met. API RP 540 and NFPA 30 set the framework; the local authority has the final word.
• Seismic and transport. Coastal sites in seismic zones need the anchorage and the tank stiffening designed for the site spectrum, and a transformer that travels by road to a remote site needs transport bracing that does not compromise the core clamping.

These are not optional refinements. A transformer ordered to a 40 °C, C3, indoor specification for a 48 °C coastal site will pass its factory tests and fail in service, usually in year four or five, with radiator corrosion or a hot-spot-related DGA trend that nobody can explain.

What does a refinery transformer cost over 20 years? The loss capitalisation arithmetic

Purchase price is often 25–40% of the lifetime cost of a transformer that runs continuously. The rest is loss. The correct way to compare offers is to capitalise the losses into the bid evaluation, using the site's own energy cost and load profile.

The method: take the no-load loss (which runs 8,760 hours a year) and the load loss (which runs at the square of the load factor), multiply by the energy price, and discount the annual cost over the asset life. At 0.08 USD/kWh, a 6% discount rate and a 20-year life, one kilowatt of no-load loss is worth about 8,000 USD, and one kilowatt of load loss at a 0.75 load factor is worth about 4,500 USD. Those two numbers turn a loss comparison into a bid adjustment in a single line.

Assumptions: 0.08 USD/kWh, 8,760 h/yr no-load, 0.75 load factor, 6% discount rate, 20-year life. Change the assumptions and the numbers change, but the ranking rarely does. The calculation is a bid-evaluation tool, not a price list.

On that basis a bid that is 80,000 USD cheaper at the purchase stage but 108,000 USD worse in capitalised loss is the more expensive transformer. The same logic applies to the choice between an amorphous-core unit and a conventional grain-oriented unit at the LV level, where no-load loss dominates because the transformer sits idle for most of its life — see our amorphous core guide if that is the decision you are making.

What belongs on the datasheet? The checklist that prevents most disputes

Most refinery transformer disputes are not about workmanship; they are about a field that was left blank or written as a range. The datasheet should carry every one of the following explicitly, with a single value or a tightly bounded range.

Two fields deserve special attention because they are the ones most often omitted and most expensive to fix later. The harmonic spectrum is usually not on the datasheet at all, which means the manufacturer sizes for sinusoidal load and the plant discovers the hotspot problem in service. The temperature rise is often left at the standard 60 K top-oil even when the site ambient is 50 °C, which quietly consumes the entire thermal margin before the first overload. If you want a single-page version of this checklist to attach to an enquiry, our nameplate reading guide explains what each of these fields means once the unit arrives, and the pre-shipment inspection checklist covers the factory stage.

What are the five most expensive specification mistakes?

• Sizing from throughput instead of a load list. It produces a rating that is either 30% too large (loss capitalisation for two decades) or too small for the largest motor start (a redesign at commissioning).
• Ignoring the largest motor start. The transformer passes every steady-state check and fails the only transient that happens several times a day.
• Leaving the harmonic spectrum off the datasheet. A drive-heavy bus derates a standard transformer by 10–25%, and the failure mode is a slow thermal one that appears after the warranty discussion has already started.
• Specifying the standard 40 °C ambient for a 48–50 °C site. The unit is compliant on paper and short-lived in practice; the fix is a lower temperature rise or an extra cooling stage, and it is nearly free at the enquiry stage.
• Treating the hazardous area classification as a civil or instrument issue. The transformer location, the fire wall, the pressurization and the cable sealing are all electrical design decisions, and relocating a substation after the plot plan is frozen is a schedule disaster.

Short answers to the questions engineers ask most

Can an oil-immersed transformer be installed in Zone 2?

Not as a certified item in the normal sense. The practical route is to place it in a non-hazardous area, or in a fire-separated room that is pressurized with air from a safe area under IEC 60079-13, with gas detection that trips the ventilation and the transformer on alarm. The certification belongs to the installation concept, not to the transformer.

Is a 2×100% substation always the right answer for a process unit?

No. It is right when the consequence of losing the unit is a plant trip, an emergency shutdown with flaring, or a safety event. For utilities and offsites, 2×50% with a bus tie or an N+1 tie to a neighbouring substation is usually the better use of capital. Decide it explicitly and record the reason, because the default answer in a FEED package is often 2×100% by habit.

How much spare capacity should the transformer have?

10–15% for load growth plus a deliberate spare feeder is normal practice; 40% blanket spare is not. The correct check is not a percentage but a load-flow study at the five-year and ten-year load, including the largest motor start at the ten-year case. Refinery debottlenecking is real, but so is the cost of carrying unused iron for twenty years.

Does a higher-impedance transformer improve safety?

It reduces fault current and therefore arc-flash incident energy, which is a genuine safety benefit, but it also worsens voltage dip and motor starting and increases regulation. The right approach is to pick the impedance that keeps the switchgear within its rating and the arc-flash energy within the site's PPE policy, then verify the largest motor start at that impedance. It is a system optimisation, not a single-number choice.

Sources / 资料来源

Source: IEC 60076-1, Power transformers — General.
Source: IEC 60076-2, Temperature rise for liquid-immersed transformers.
Source: IEC 60076-3, Insulation levels, dielectric tests and external clearances in air.
Source: IEC 60076-5, Ability to withstand short circuit.
Source: IEC 60076-7, Loading guide for oil-immersed power transformers.
Source: IEC 60076-11, Dry-type transformers.
Source: IEC 60079-10-1, Classification of areas — Explosive gas atmospheres.
Source: IEC 60079-13, Equipment protection by pressurized room “p” and artificially ventilated room “v”.
Source: IEC 60079-14, Electrical installations design, selection and erection.
Source: IEEE C57.12.00, General requirements for liquid-immersed distribution, power and regulating transformers.
Source: IEEE C57.110, Recommended practice for establishing transformer capability when supplying non-sinusoidal load currents.
Source: IEEE 519-2022, Recommended practice and requirements for harmonic control in electric power systems.
Source: IEEE 1584-2018, Guide for performing arc-flash hazard calculations.
Source: IEEE 399, Recommended practice for industrial and commercial power systems analysis.
Source: API RP 540, Electrical installations in petroleum processing plants.
Source: API RP 500, Classification of locations for electrical installations at petroleum facilities.
Source: API RP 505, Classification of locations for electrical installations at petroleum facilities classified as Zone 0, Zone 1 and Zone 2.
Source: NFPA 70 (National Electrical Code), Articles 450, 500 and 505.
Source: NFPA 30, Flammable and combustible liquids code.
Source: ISO 12944, Corrosion protection of steel structures by protective paint systems.
Source: GB/T 6451, Technical parameters and requirements for oil-immersed power transformers.
Source: GB 20052, Minimum allowable values of energy efficiency and energy efficiency grades for power transformers.
Source: QDTB engineering experience — refinery substation sizing, motor-starting and harmonic-derating practice, industry experience values.

If you are preparing an enquiry for a refinery or petrochemical substation, the fastest way to a comparable offer is to send the load list, the single-line diagram, the hazardous area drawing and the ambient and corrosion data together. Our engineering team can turn that into a sized transformer specification and an indicative budget in one pass; you can also start from the power system calculators to check the fault level, the voltage dip and the loss capitalisation before you talk to any supplier.