LNG Terminal Transformer: How to Size the 33 kV Supply for a 4 mtpa Regas Terminal, Why Send-Out Pumps and 2 × 100% Duty Set the Spec

The answer first: a 4 mtpa regasification terminal is a 21,423 kW, 23.3 MVA load, and two 31.5 MVA 33/6.6 kV units carry it — 2 × 100%, not N+1

Build the number from the pump curve, not from a rule of thumb: 4.0 million tonnes a year of LNG is 10,959 tonnes a day, which at 48 GJ per tonne is 53.3 TWh of send-out thermal energy a year, 6.09 GW average and roughly 7.0 GW on a peak winter day. Moving that much liquid methane from a tank at 1.1 bar to a pipeline at 80 bar is a pumping duty, not a compression duty, and it draws 21,423 kW at the unloading peak — 23.3 MVA at 0.92 power factor. Two 31.5 MVA 33/6.6 kV ONAN/ONAF units with 10.5% impedance serve it: 37% loaded in normal operation, 74% of the ONAN rating with one unit out, which is what firm send-out capacity actually requires. The redundancy argument at an LNG terminal is not the factory argument. A terminal signs a firm capacity commitment with a gas buyer and pays take-or-pay if it under-delivers, so it buys 2 × 100% transformers and a bus tie that closes in 8 seconds, and it spends the money on 18-pulse VFD front ends rather than on a third transformer. Three things separate a terminal specification from a plant of the same kVA: the load is 85% large VFD motors on one 6.6 kV bus, so harmonic distortion rather than kVA decides the transformer impedance; the site is a marine environment at C5-M corrosivity with 31 mm/kV creepage, so bushings and paint cost more than the core grade; and the plant runs 8,760 hours a year at 30% average loading, which makes no-load loss worth ten times load loss and rewrites the loss-evaluation formula in the tender.

Key takeaways

• 4.0 mtpa = 10,959 t/day, 502 t/hour at peak, 6.09 GW average send-out. The electrical load is 26,120 kW connected and 21,423 kW at the unloading peak, 23.3 MVA at 0.92 PF; the send-out pumps alone are 11,970 kW of that peak.
• Two 31.5 MVA 33/6.6 kV ONAN/ONAF units at 37% normal and 74% single-unit loading, with a 6.6 kV bus tie and 8-second automatic transfer, meet the firm capacity commitment. A third transformer adds USD 420,000 of capital and 0.02% of availability.
• Impedance and K-factor are set by harmonics, not by voltage regulation: 12 six-pulse VFDs on one bus give 30% current THD, 18-pulse front ends bring TDD to 4.1% at the 6.6 kV bus and 2.8% at the 33 kV point of common coupling, inside the 8.0% that IEEE 519-2022 allows in the 10–69 kV band.
• At 30% average loading, no-load loss is worth USD 701 per kW per year and load loss only USD 68 per kW per year. The two S11 units burn USD 56,200 a year; S13 cores save USD 13,300 a year for about USD 12,000 of capital, and amorphous cores pay back in under four years — the opposite conclusion to a hospital, and for the same arithmetic reason.
• Marine environment, not altitude, drives the materials: ISO 12944-2 C5-M paint at 320 µm, IP55 enclosures, 31 mm/kV creepage, silicone composite bushings, stainless hardware and anti-condensation heaters are specification items that appear on every LNG terminal datasheet and on almost no factory datasheet.

What does a 4 mtpa regasification terminal actually draw? The load schedule behind the kVA number

A regasification terminal is often described as a gas plant, which makes electrical engineers expect a compression load with a large synchronous motor and a soft starter. That description is wrong for a modern send-out system. The cheap way to raise LNG from atmospheric pressure to 80 bar is to do it in the liquid phase, where the density is 450 kg/m³ and the volume is 1/600 of the gas volume, and then to vaporise the pressurised liquid in an open-rack vaporiser using seawater. Compression of 6 GW of gas would need something like 150 MW of compressors; pumping the same energy as liquid needs 12 MW. Every terminal built in the last twenty years takes the pumping route, and the electrical consequence is that the plant is a pump plant: four large high-pressure send-out pumps, six in-tank low-pressure pumps, two boil-off gas compressors and four seawater pumps, all on a 6.6 kV bus, all variable speed.

The load schedule below is the single document that decides the transformer. It is built the way a terminal EPC contractor builds it: by equipment list, with a diversity factor applied per group, and with the peak defined as the coincidence of ship unloading, full send-out and the seawater system at design ambient. The unloading coincidence matters because a 4 mtpa terminal receives about 60 cargoes a year, a 145,000 m³ cargo every six days, and unloads it in 26 hours while send-out continues.

Two numbers in that table are worth arguing about, because they are where terminal specifications usually go wrong. The first is the diversity factor on the high-pressure pumps: 0.95 is not a diversity factor at all, it is a statement that during unloading all four pumps run at 95% of rated shaft power because the send-out rate is at its contractual maximum and the tanks are being filled at the same time. Designers who apply 0.8 to that row and then add 10% spare capacity arrive at 18 MW, buy two 20 MVA transformers, and discover on the first winter peak that the plant cannot hold 100% send-out with one unit out. The second is the boiler-off gas compressor row, where the load is set by the unloading vapour displacement — 10,000 m³/h of liquid entering the tank displaces about 12,000 m³/h of vapour, which is 5.4 t/h of methane — and not by the 0.05%/day tank boil-off, which is only 229 kg/h and would justify a 250 kW compressor rather than a 2,200 kW one.

With 21,423 kW at 0.92 power factor the demand is 23,286 kVA. Add the 5% design margin that every terminal specification carries for future debottlenecking and the target becomes 24.5 MVA, which is why 25 MVA looks like the natural size and why 31.5 MVA is the size that actually gets bought. The difference is redundancy, and it is the subject of the next section.

Why 2 × 100% and not N+1: what firm capacity does to transformer sizing

In a factory, a transformer failure costs production until the spare arrives, and the plant accepts that because the spare is expensive and the loss is measurable in shifts. In an LNG terminal, a transformer failure costs the difference between the contracted send-out and the delivered send-out, priced at the buyer's alternative supply cost, and the terminal usually carries a take-or-pay obligation on the regasification capacity it has sold. That changes the economics from “a spare transformer is a luxury” to “a spare transformer is the cheapest insurance in the plant”, and it has a specific consequence for sizing: the standby must be able to carry 100% of the peak, not 100% of the normal load.

That is the whole argument, and it explains why a 23.3 MVA peak produces 2 × 31.5 MVA rather than 2 × 25 MVA. Two 25 MVA units would each sit at 47% in normal operation and at 93% with one out — which is inside the ONAN rating at 45 °C ambient only if the unit is derated by 8% for temperature, leaving 23 MVA, and the peak is 23.3 MVA. The arithmetic fails by 1.3%, which on the drawing becomes a note about load shedding during unloading, which becomes a contractual argument with the gas buyer. Two 31.5 MVA units, derated to 29 MVA at 45 °C, carry the peak at 74% with one unit out and leave room for the debottlenecking margin. The extra capital is about USD 60,000 for the pair at this size; the alternative is a load-shedding scheme and a permanent argument.

The bus tie is where the redundancy is actually implemented, and it is worth being precise about the switching sequence because it decides the transformer impedance and the protection settings. The 6.6 kV switchboard is split into two sections, each fed by one transformer, with a normally open tie breaker. On loss of one incomer, the tie closes after a 1.5-second confirmation delay plus the motor residual-voltage decay, and the surviving transformer picks up the whole bus in about 8 seconds including the VFD restart ramp. That sequence is why the transformers are specified with 10.5% impedance rather than 8%: with 8% impedance the fault level at 6.6 kV rises from 26.2 kA to 34.4 kA, the switchgear moves from 31.5 kA to 40 kA, and the extra switchgear cost exceeds the loss saving from the lower impedance. It is also why parallel operation of the two transformers through a closed tie is not permitted as a normal configuration: the circulating current at 10.5% impedance and a 1% tap mismatch would be about 300 A, and the 6.6 kV fault level would double.

The control philosophy completes the picture. Send-out pumps are grouped in two trains, each train fed from one 6.6 kV section, and each train is sized for 60% of peak send-out. Losing a transformer loses one train, which leaves 60% send-out for the 8 seconds until the tie closes — a controlled ramp rather than a trip, and a reduction that the pipeline operator tolerates under the operating agreement. That is a design decision worth stating explicitly in the electrical philosophy document, because it is what makes 2 × 100% transformers with a bus tie equivalent to 3 × 50% with no tie, at 60% of the switchgear cost.

Harmonics: why 12 six-pulse VFDs decide the transformer impedance and the K-factor

By connected load, 74% of the terminal is variable-speed drive: 12.6 MW of send-out pumps, 4.4 MW of boil-off gas compressors, and the seawater pumps if they are also on drives. A six-pulse diode front end draws a current waveform with 25% fifth harmonic, 12% seventh, 6.5% eleventh and 4.5% thirteenth, for a current THD of about 30% at each drive. Twelve drives on one bus do not add arithmetically — the fifth and seventh components partly cancel because the drives have different loading and the phase angles drift — but the measured result at a terminal bus with 12 six-pulse drives is typically 14–18% current THD, not 30% and not 4%. That is well above the 8.0% total demand distortion that IEEE 519-2022 allows in the 10–69 kV band at the point of common coupling with an Isc/IL ratio around 30, and it is the reason the harmonic study, not the load flow, fixes the transformer specification.

There are three ways to meet the limit, and the choice is economic rather than technical. The first is to buy drives with 18-pulse front ends, which use a phase-shifting transformer or a triple-winding input reactor to cancel the fifth and seventh; the result is 4.1% TDD at the 6.6 kV bus and 2.8% at 33 kV, at a drive premium of 12–18%. The second is an active front end, which gives 3.5% TDD and regeneration capability but costs 25–35% more and adds a source of high-frequency emissions that has to be filtered for the plant's own instrumentation. The third is a passive fifth and seventh filter on the 6.6 kV bus, which is the cheapest at 6.8% TDD but has a resonance that interacts with any power factor correction capacitors and with the transformer's own impedance, and which stops working when the drive loading changes.

The transformer consequences are two, and they are frequently missed in the datasheet. The first is that harmonic currents raise the winding eddy-current loss, and the standard treatment is a K-factor or a harmonic loss factor; for the 18-pulse case the K-factor is 4, and for an unfiltered six-pulse bus it is 13. Specifying a standard distribution transformer with K-1 in a terminal is a common and expensive mistake: the hot spot moves from the winding surface into the conductor, the 65 K temperature rise becomes 78 K, and the insulation ages at twice the rate. If the bus cannot be cleaned up, the transformer has to be specified with a K-factor, and the reference articles on K-factor transformer selection and transformer harmonics explain how the factor is calculated rather than guessed.

The second consequence is impedance. The harmonic currents flow through the transformer's leakage reactance and produce a voltage distortion at the 6.6 kV bus that is proportional to the impedance and to the harmonic order — the fifth harmonic sees five times the reactance. At 10.5% impedance and 4.1% TDD, the fifth-harmonic voltage distortion at the bus is about 1.8%, which is inside the 3% individual limit of IEEE 519; at 8% impedance it is 1.4%, and the difference is not worth the 40 kA switchgear. What does matter is the interaction with any capacitor bank: the resonant frequency is f_r = f × √(S_sc/Q_c), and a 2 MVA capacitor bank on a 300 MVA bus puts the resonance at the 12th harmonic, which is exactly where the 18-pulse drive leaves its residual. That is why terminals with VFDs either avoid power factor correction capacitors altogether and let the drives' own front ends hold the power factor at 0.96, or install detuned reactors at 7% — and why the specification should say which, because the transformer's impedance is chosen with that decision already made.

Motor starting and voltage dip: what the VFDs are really for

Every large drive at a terminal is on a VFD, and the reason usually given is process control — send-out rate has to follow the pipeline nomination, which changes hourly and sometimes by 20% in a minute. But the electrical reason is starting current. A 3,150 kW 6.6 kV motor has a full-load current of 287 A and a direct-on-line starting current of six to six and a half times that, 1,865 A, which is 21.3 MVA of starting demand. On a 6.6 kV bus fed by one 31.5 MVA transformer with 10.5% impedance, the source impedance is 300 MVA, so the dip at the bus during a DOL start is 21.3 divided by 321.3, or 6.6%. That is just inside the 10% that most terminals allow for a single start, and it is outside the 5% that the plant's own instrumentation and the marine loading arm drives will tolerate, which is the practical reason the drives are variable speed.

Two details matter for the transformer beyond the dip calculation. The first is that a VFD is a constant-power load above base speed, so the current stays at rated while the power rises; the transformer sees a load with a displacement power factor of 0.96 but a true power factor of 0.91 because of the distortion, and the kVA rating has to be evaluated on the true power factor. The second is the VFD's own contribution to the fault current: an 18-pulse front end with a DC link limits its fault contribution to about 1.5 times rated current for one cycle, but 12 of them on one bus still add 4 kA to the 26.2 kA transformer contribution, which is why the switchgear is rated 31.5 kA rather than 25 kA and why the transformer's short-circuit withstand is specified at 25 kA for 2 seconds per short-circuit current calculation rather than the 20 kA that a 25 MVA unit would need.

The inrush point is worth a sentence because terminals trip on it. Energising a 31.5 MVA 33/6.6 kV transformer produces an inrush of 8–12 times rated current for 0.1 to 0.3 seconds, with a DC offset that can saturate the core for several cycles. On a bus with 12 VFDs, the VFDs see the voltage sag and, depending on their under-voltage ride-through setting, either ride through or trip. The fix is a controlled switching relay that closes at the voltage zero crossing, which cuts the inrush to 2–3 times rated and is worth the USD 8,000 it costs on a transformer of this size; the alternative is to accept a 15% dip and a VFD restart sequence every time the plant is re-energised. The mechanics are covered in transformer inrush current.

Marine corrosion, creepage and the materials that an LNG terminal specification actually buys

A terminal is built on a jetty or on reclaimed land next to salt water, and the transformer sits outdoors in an atmosphere that ISO 12944-2 classifies as C5-M: chloride deposition above 300 mg/m² per day, 90–95% relative humidity, and a salt fog that never dries off completely because the night humidity is above the deliquescence point. That environment does not change the transformer's electrical design, but it changes every external material, and the cost of getting it wrong shows up three years after commissioning as a bushing flashover or a rusted tank seam.

The items that have to be on the datasheet, and the reason for each, are these. The paint system is a three-coat epoxy-polyurethane at 320 µm total dry film, specified to ISO 12944-5 for C5-M with a 15-year durability band; the standard 120 µm alkyd that comes with a distribution transformer fails at the weld seams within 24 months in a coastal terminal. The enclosure for the marshalling box, the OLTC cabinet and the control cabinet is IP55 rather than IP54, with stainless steel 316 hinges and A4-70 fasteners, because galvanised hardware becomes a rust streak and then a seized bolt. Bushings are silicone composite rather than porcelain, with a creepage of 31 mm/kV — that is 31 mm per kV of line-to-line voltage, so 1,023 mm for a 33 kV bushing — because the 20 mm/kV porcelain that is standard for inland sites flashes over in salt fog at about 60% of its dry withstand. The tank is painted with a stripe coat on every weld and the radiator banks are hot-dip galvanised after fabrication rather than painted, because paint on a radiator fin cannot be maintained.

There is a second environmental item that is easy to miss: condensation. A transformer that runs at 37% load has a tank top that is cooler than the humid night air for several hours a day, and the oil conservator breathes that air in through the silica gel. Terminals in the tropics specify a sealed tank with a nitrogen blanket or a conservator with a membrane, plus anti-condensation heaters in every cabinet and terminal box, and a dew-point alarm on the breather. The alternative is water in the oil, which shows up as a dissolved gas analysis result two years later and as a failed oil test at the first annual service. The condition-monitoring approach is described in the reference material on transformer maintenance, but the specification point is simpler: buy the sealed design at the factory, because retrofitting it on site costs three times as much.

Ambient temperature and solar radiation complete the derating. Design ambient is 45 °C with a 50 °C daily peak in the shade and 1,000 W/m² of solar radiation on the tank, which for an ONAN unit adds about 4 K to the top-oil rise. The IEC 60076-7 loading guide gives the derating: at 45 °C ambient, a unit rated 31.5 MVA at 40 °C delivers 29.0 MVA at 65 K rise, and 32.4 MVA with the fans running (ONAF). That 29.0 MVA number, not the 31.5 MVA nameplate, is what the redundancy calculation in the previous section uses, and it is why the datasheet has to state the ambient and the cooling stage explicitly rather than leaving it to a note.

Hazardous area, fire separation and where the transformer is allowed to stand

LNG is methane, and methane is flammable, but a well-designed terminal has very little of it outside the pipework. The hazardous area classification per IEC 60079-10-1 puts the boil-off gas compressor house, the vapour return connection and the area within 15 m of an LNG flange as Zone 2, with Zone 1 inside the compressor enclosure and at the relief valve outlets. An oil-immersed transformer is not permitted inside a Zone 2 area at all, not because it is a source of ignition in normal operation — a sealed tank is not — but because NFPA 59A and EN 1473 treat a large oil volume as a fire risk that has to be separated from a gas release, and because a transformer fire next to a gas leak is the one scenario a terminal's quantitative risk assessment will not accept.

The practical consequence is that the main substation is located 30 m from the compressor house and outside the classified area, with the 33 kV and 6.6 kV switchgear in a pressurised electrical building, and the cable route to the process area is either in a buried duct bank or in a pressurized corridor. Where a transformer has to be close to the process — a 2,500 kVA 6.6/0.4 kV unit for a compressor house, for example — the specification calls for a cast-resin dry-type unit, which has no oil and is accepted in Zone 2 with a 3 m separation. That is the one place where dry-type wins at a terminal, and it is worth knowing because the general rule at an LNG plant is the opposite of a building: oil-immersed outdoors for the big units, cast resin only where the hazardous area forces it.

The fire protection around the main transformers is specified by NFPA 59A and by the insurer, and it usually consists of a 2-hour fire wall between the two units at 1.5 m from the tank, an oil containment pit sized for 110% of the oil volume of the largest unit, a deluge system at 10 l/min/m² over the tank and radiators, and a foam injection point. The separation between the two transformers matters for the redundancy claim: at 1.5 m with a fire wall the second unit survives a fire in the first, which is the event the 2 × 100% arrangement is meant to cover. Terminals that put the two units 6 m apart without a wall have the capital cost of 2 × 100% and the fire resilience of N+1, and the insurance survey usually finds it.

Earthing and lightning: the pipe rack is the electrode

A terminal's earthing system is unusual because the plant itself is a large earthed structure. The LNG pipework, the tank shells, the jetty steel and the marine loading arms are all bonded, and the earth grid is a 70 mm² bare copper ring around the process area with 25 mm × 3 mm copper tapes to every structural column, sized for a 25 kA fault for 1 second and for a lightning strike of 200 kA at 10/350 µs per IEC 62305. The design target is 1 Ω at the main earth bar, measured at commissioning with a fall-of-potential test, and the step and touch voltages are checked to IEEE 80 at the switchyard rather than at the process area.

Two terminal-specific items are worth putting in the specification. The first is cathodic protection interference: the jetty and the marine structures have impressed-current cathodic protection, and the earth grid is bonded to them through DC decoupling spark gaps so that the CP current does not flow through the copper grid and corrode it. An LNG terminal that bonds the grid directly to a CP-protected jetty loses the jetty anodes within two years and finds the problem as a coating failure, not as an electrical fault. The second is the surge protection coordination: 33 kV line entrance arresters, 6.6 kV arresters at the transformer terminals, and Type 2 SPDs at every LV board, coordinated by cable length so that the 33 kV surge is attenuated before it reaches the 6.6 kV winding. The insulation coordination is the same calculation as any substation, covered in transformer lightning protection and low-voltage grounding systems, but the marine bonding and the CP decoupling are the parts that inland specifications leave out.

Tap changer and voltage regulation on a weak 33 kV grid

Terminals are often built at the end of a long 33 kV or 132 kV line, next to a gas pipeline and a coastline, in a place where the grid is a single radial feeder with a short-circuit level of 800–1,500 MVA and a voltage that moves by 8% between the peak and the light-load night. The variable-speed drives tolerate ±10% voltage without derating, so the process does not need tight regulation, but the 6.6 kV bus has to stay inside ±5% for the LV transformers and the plant's own electronics. That is a case for an on-load tap changer on the 33 kV winding, ±8 steps of 1.25%, with an automatic voltage relay measuring the 6.6 kV bus and a time delay of 60 seconds to ride through motor starts.

The alternative — an off-circuit tap changer set at commissioning — is cheaper by USD 25,000 per unit and is what a factory would buy. At a terminal it is the wrong choice for a specific reason: the send-out nomination follows the gas price, and a terminal that runs at 40% send-out for a month and then at 100% for a week sees the bus voltage move with the load because the transformer's own regulation is 4–5% at 10.5% impedance. An OLTC fixes that automatically; a fixed tap means the operator changes it twice a year, in a shutdown, with the transformer isolated. The reference material on on-load versus off-load tap changers covers the mechanism; the terminal-specific point is that the OLTC is justified by the load profile rather than by the grid.

Losses at 8,760 hours: why the tender's loss-evaluation formula is usually wrong

Every transformer tender carries a loss-evaluation formula that converts no-load and load loss into a capital equivalent, and at an LNG terminal that formula is almost always copied from a power utility where transformers run at 60–70% average loading. A terminal runs at 30% average loading for 8,760 hours a year, so the two loss components have very different values, and getting the weights wrong buys the wrong core.

The arithmetic is simple once the load profile is known. Annual average demand is 18.3 MVA, which is 29% of the installed 63 MVA, so the load loss at average load is 0.29² = 8.4% of the rated load loss, and the equivalent full-load hours for load loss are 0.29² × 8,760 ≈ 737 hours, call it 850 with the peak weeks included. No-load loss runs for all 8,760 hours. At USD 0.08/kWh, no-load loss is worth USD 701 per kW per year and load loss USD 68 per kW per year — a ratio of ten to one. A tender that values load loss at USD 2,500/kW and no-load loss at USD 5,000/kW, which is a typical utility formula, is undervaluing no-load loss by 40% relative to its true cost at this duty.

The conclusion for this duty is unusual and worth stating plainly: at 8,760 hours and 30% loading, amorphous cores do pay back, in under four years, because no-load loss dominates. That is the opposite of the conclusion for a hospital of similar size, where the same arithmetic with 4,050 equivalent hours makes the amorphous premium uneconomic. The difference is not the transformer; it is the number of hours the core is energised and the price of the energy. A terminal that buys S11 cores to save USD 28,000 of capital spends about USD 13,300 a year more on electricity for the life of the plant, which at a 7% discount rate over 25 years is about USD 155,000 of present value — five times the saving. The reference material on loss grades S11, S13 and S20 covers the grade definitions, and the total-cost method is set out in transformer total cost of ownership.

What the datasheet has to say: a specification table for a terminal transformer

The specification below is the one that a terminal EPC package produces after the studies in the previous sections, and it is worth comparing with the datasheet of a distribution transformer of the same rating because almost every line differs. The items in bold are the ones that are specific to an LNG terminal and that a standard quotation will not include unless they are asked for.

Transport, installation and commissioning: the schedule risk that is not electrical

A 31.5 MVA 33/6.6 kV transformer is about 55 tonnes with oil, 6.4 m long, 2.6 m wide and 3.6 m high, and it moves from a factory in China to a jetty in West Africa or Southeast Asia on a flat rack, then by road on a multi-axle trailer at 40 km/h, then onto a plinth 30 m from the compressor house. The electrical design is settled by then; the schedule risk is the transport, and the two items that protect it are an impact recorder with a 3 g threshold on the tank and a nitrogen fill at 0.02 MPa for the sea voyage, both of which are in the reference material on transformer transport and installation.

At site, the sequence is: receive and inspect against the impact recorder, stand the unit on the plinth with the jacking pads, vacuum-dry if the oil was drained for transport, oil-fill under vacuum to 1 mbar, soak for 24 hours, then the commissioning tests. The tests that matter at a terminal are the winding resistance (which finds a loose connection before it becomes a hot spot), the ratio and vector group (which finds a reversed CT or a wrong tap), the insulation resistance and tan delta at 10 kV for the 6.6 kV winding, the SFRA signature compared with the factory baseline, and the OLTC through its full range with the motor drive. The reference checklist for this sequence is in commissioning tests, and the pre-shipment inspection that protects the schedule is in pre-shipment inspection.

Two terminal-specific commissioning items are easy to forget and expensive to retrofit. The first is the harmonic measurement at the 6.6 kV bus under real load, which cannot be done until the send-out pumps run, so it belongs in the performance test rather than in the commissioning schedule; a terminal that discovers 9% TDD instead of 4.1% after handover is looking at a filter retrofit in a live plant. The second is the OLTC and AVR settings under a real grid voltage swing, which requires the grid operator to move the tap on the feeding transformer, and which is usually deferred and then forgotten until the first voltage complaint.

What the electrical package costs, and where the money actually goes

The figures below are industry experience ranges for a 4 mtpa terminal's main electrical package, FOB China, 2025–2026, excluding installation and marine works. They are given as ranges because the spread between a tier-one European supplier and a Chinese supplier of the same specification is 35–50%, and because the transformer is a smaller share of the package than most project managers expect.

The two transformers are 19% of the package; the drives are 29%. That ratio is the practical argument for spending engineering effort on the harmonic study rather than on the core grade: a decision that changes the drive front end moves USD 200,000, while a decision that changes the core grade moves USD 30,000. It is also the argument for buying the transformers from a supplier who will quote the C5-M paint, the 31 mm/kV bushings and the K-factor as line items rather than as “marine execution” at a lump sum, because that is where the 35–50% spread between suppliers is really hiding.

Seven specification mistakes that show up at every LNG terminal

• Applying a factory diversity factor to the send-out pumps. 0.8 instead of 0.95 gives 18 MW instead of 21.4 MW, two 20 MVA transformers instead of two 31.5 MVA, and a plant that cannot hold 100% send-out with one unit out — the exact event the redundancy was bought for.
• Specifying K-1 transformers on a VFD bus. The harmonic loss factor at 30% current THD is 1.15 and at 4.1% TDD is 1.08; a standard unit runs its hot spot 10–13 K hotter and ages at twice the rate. The K-factor costs 3% of the transformer price.
• Buying 25 MVA because 23.3 MVA fits. The derated rating at 45 °C ambient is 29.0 MVA for a 31.5 MVA unit and 23.0 MVA for a 25 MVA unit, so the 25 MVA pair is 100.1% loaded on the peak day with one unit out.
• Leaving the corrosion class at C3. Standard paint at 120 µm fails at weld seams within two years in a C5-M atmosphere; the repaint at height, over a live terminal, costs more than the original C5-M system.
• Porcelain bushings at 20 mm/kV. They pass the factory impulse test and flash over in salt fog at 60% of dry withstand; silicone composite at 31 mm/kV is the fix, and it must be specified, not assumed.
• Using a utility loss-evaluation formula. At 8,760 hours and 30% loading, no-load loss is worth ten times load loss; a formula weighted the other way buys S11 cores and wastes about USD 13,300 a year.
• Placing the substation inside the Zone 2 boundary. Oil-immersed units are excluded by NFPA 59A and EN 1473; the correction after the HAZOP is a 30 m cable route and a redesign of the 6.6 kV bus.

Sources / 资料来源

Source: IEC 60076-1:2011 — Power transformers, Part 1: General.
Source: IEC 60076-2:2011 — Temperature rise for liquid-immersed transformers.
Source: IEC 60076-3:2013 — Insulation levels, dielectric tests and external clearances in air.
Source: IEC 60076-5:2006 — Ability to withstand short circuit.
Source: IEC 60076-7:2018 — Loading guide for mineral-oil-immersed power transformers.
Source: IEC 60076-11:2018 — Dry-type transformers.
Source: IEC 60137:2017 — Insulated bushings for alternating voltages above 1,000 V.
Source: IEC 60815-1:2008 — Selection and dimensioning of high-voltage insulators intended for use in polluted conditions.
Source: IEC 60909-0:2016 — Short-circuit currents in three-phase AC systems.
Source: IEC 60079-10-1:2020 — Classification of areas: explosive gas atmospheres.
Source: IEC 62305-3:2010 — Protection against lightning, physical damage to structures.
Source: IEEE 519-2022 — Recommended practice and requirements for harmonic control in electric power systems, Table 2 limits for 10–69 kV.
Source: IEEE 80-2013 — Guide for safety in AC substation grounding.
Source: IEEE C57.12.00-2021 — General requirements for liquid-immersed distribution, power and regulating transformers.
Source: NFPA 59A — Standard for the production, storage, and handling of liquefied natural gas.
Source: NFPA 20 — Standard for the installation of stationary pumps for fire protection.
Source: EN 1473:2021 — Installation and equipment for liquefied natural gas: design of onshore installations.
Source: ISO 12944-2:2017 and ISO 12944-5:2019 — Corrosion protection of steel structures, corrosivity categories and paint systems.
Source: GIIGNL Annual Report 2025 — global LNG trade volumes and regasification capacity.
Source: Industry experience — send-out pump hydraulic duties, boil-off gas rates and unloading vapour displacement for 3–5 mtpa onshore regasification terminals in Southeast Asia, the Middle East and West Africa, 2018–2025.
Source: Industry experience — measured current TDD of 14–18% on 6.6 kV terminal buses with 10–14 six-pulse drives, and 3.5–4.5% with 18-pulse front ends, from harmonic surveys at commissioned terminals, 2019–2025.
Source: Industry experience — transformer, GIS, switchgear and VFD price ranges for LNG terminal electrical packages, FOB China, 2025–2026.
Source: Industry experience — loss-evaluation values of USD 400–1,100 per kW and industrial electricity tariffs of USD 0.07–0.10 per kWh used in the annual loss-cost calculation.
Source: QDTB Engineering Team — load schedules, harmonic studies and transformer specifications for LNG, gas processing and marine terminal projects in Southeast Asia, the Middle East and Africa, 2018–2026.