Desalination Plant Transformer: How to Size the MV Substation for an SWRO Plant, Why VFD-Driven High-Pressure Pumps Decide the Harmonics and the kVA, and What Coastal Corrosion Demands

The answer first: a desalination plant transformer is sized from the RO train's pump duty and a documented harmonic study, not from a kVA-per-cubic-metre rule of thumb. In a seawater reverse osmosis (SWRO) plant the high-pressure feed pumps and their variable-frequency drives (VFDs) are at the same time the largest load and the largest distortion source, a handful of very large induction motors set the worst-case voltage dip at the pumping busbar, and a coastal site sets the corrosion and ingress class of every transformer and enclosure. The practical sequence is therefore: build the load list process area by process area, apply demand and diversity, convert to kVA, check motor-starting dip at the high-pressure pump bus, derate the transformer for the non-sinusoidal load current it will actually carry, and only then apply ambient and corrosion derating before a standard rating is selected.

A desalination plant transformer is the unit that takes the voltage a utility delivers to a coastal site — commonly 33 kV, 110 kV or 132 kV, and 220 kV or 400 kV at very large plants — and turns it into the electricity that actually moves water: 11 kV or 6.6 kV for intake, high-pressure and booster pumps driven by medium-voltage VFDs; 11 kV or 6.6 kV down to 0.4 kV or 0.69 kV for pre-treatment, dosing, backwash, clean-in-place (CIP) and low-voltage motor control centres; and 0.4 kV auxiliaries for instrumentation, electrical-room HVAC and lighting. What separates desalination duty from an ordinary utility distribution load is not the total kVA. It is that one process step — the high-pressure RO feed pump — dominates the load and cannot be interrupted, that most of that load arrives through a power-electronic rectifier, and that the environment is marine. Those three facts drive the sizing, the harmonic specification and the enclosure, in that order.

Everything below assumes one thing: the transformer sits inside a water factory whose output is sold at a fixed price per cubic metre. Unavailability is therefore measured in lost m³, not in lost kVA, and that is why desalination electrical design leans on N-1 redundancy, on harmonic margins and on voltage-dip limits that a utility distribution feeder would never need. If you are sizing the equipment on the other side of the meter, the general method in our sizing and budget formula guide applies; this article covers what changes when the load is an RO train on a coast.

Key takeaways

• Size from the water, not from a table. Convert required product flow (m³/day) and recovery into feed flow and pressure, then into hydraulic power, then into electrical kVA through pump and motor efficiency. A 100,000 m³/day SWRO plant lands near a 20 MVA firm load — the worked example below shows every step.
• The load is converter duty, not sine-wave duty. A six-pulse rectifier draws 30–45% current THD; specify the transformer for converter duty against IEC 61378-1 and derate it for non-sinusoidal load to IEEE C57.110, exactly as in our K-factor transformer selection guide.
• Motor starting decides the impedance, not the kVA. Direct-on-line starting of a 2,000 kW 6.6 kV motor draws about 6 × full-load current; check the dip at the pump busbar before you accept a transformer impedance — see short-circuit current calculation.
• Coastal corrosion is a specification line, not a paint choice. ISO 12944 C5-M (or CX in the splash zone) coating systems, hot-dip galvanising to ISO 1461, IP54 or better to IEC 60529 and 316 stainless hardware belong in the transformer order, together with the oil-immersed vs dry-type decision that follows from indoor or outdoor location.
• N-1 must be designed at every transformation stage, with a bus-tie and a load-shedding scheme, not merely by buying two transformers — the same conditions that govern parallel operation apply at every tier.

What electrical architecture does a seawater desalination plant use?

Modern SWRO plants almost always use a two-stage transformation: a main receiving substation where the utility voltage steps down to an internal distribution voltage (33 kV is the classic choice, 6.6 kV and 11 kV also appear at smaller plants), and satellite substations at each process area that step 33 kV down to 11 kV or 6.6 kV next to the pumps that use it. The intermediate 33 kV tier is not decoration: it carries bulk power across a site that is physically spread out — intake on the shoreline, pre-treatment, RO building, product-water pumping, outfall — at a current and cable size that stays manageable, and it hands power down to the motor voltage only where the motors are.

Attempting a single 33/11 kV (or 33/6.6 kV) step for a large plant does not work in practice, because the low-voltage-side busbar current becomes enormous and the load centres are hundreds of metres apart. That is the same cable-versus-voltage logic that forces a higher voltage in mining pit distribution and in the 33 kV collector systems of utility-scale solar and wind farms.

Two features distinguish this architecture from a normal industrial plant of the same size. First, the energy flows are concentrated in a single mechanical step: the high-pressure pump train, which typically accounts for 60–75% of the plant's electrical demand. Second, the process cannot be restarted as cheaply as it can be stopped: an uncontrolled trip of the intake or high-pressure system risks membrane fouling, pressure-vessel damage and hours of re-stabilisation, so the electrical design is built around availability rather than around first cost.

Why do the high-pressure pumps decide the transformer specification?

The RO high-pressure pump must raise the full feed flow to the osmotic-plus-net-driving pressure of the membranes — typically 55–70 bar for seawater, higher for cold feed water and for high-recovery designs. Recovery is usually 40–50% for a single-pass seawater train, so the feed flow is roughly twice the product flow and the hydraulic power is set by the product of that flow and that pressure. Three consequences follow directly for the transformer:

1. The load is large, few and lumpy. A 100,000 m³/day plant runs 3–5 high-pressure pumps of 1.5–3 MW each rather than hundreds of small motors. A single transformer therefore sees a load that steps in large increments, and the loss of one drive is a large step in busbar voltage — the reason voltage regulation and tap range matter (voltage regulation explained).
2. Flow control is done electrically, not by throttling. Because membranes must not be throttled like a valve, flow and pressure are trimmed with a VFD or with a combination of VFD and a hydraulic energy-recovery device. That puts a rectifier between the transformer and the motor.
3. Motor-starting kVA is a transformer design input. Even when the main high-pressure pumps are VFD-driven, the intake, backwash, transfer and outfall pumps are often direct-on-line or soft-started, and their starting current shares the same busbar. Whether the starting dip is acceptable is decided by the transformer impedance and the fault level at that busbar, not by the drive.

Specific energy consumption (SEC) — kWh per cubic metre of product water — is the number that ties all of this together. Best-in-class SWRO plants now report roughly 2.0–3.0 kWh/m³ for the RO loop with a modern energy-recovery device, and a verified figure of 1.794 kWh/m³ at 40% recovery has been achieved on a 2,500 m³/day plant; older and larger plants frequently sit in the 3.0–4.5 kWh/m³ band once intake, pre-treatment and product-water pumping are included. Those ranges are the reason a desalination plant is a 24-hour, high-load-factor customer: unlike a factory that runs two shifts, the plant wants to run flat out all year, which pushes the transformer towards continuous full-load duty and makes efficiency and cooling choices pay back quickly (efficiency standards, loss grades compared).

How do VFD harmonics change transformer sizing and specification?

A diode or thyristor rectifier does not draw a sine wave. It draws current in short pulses near the voltage peaks, so the current on the transformer secondary contains characteristic harmonics of order 6k ± 1 — the 5th, 7th, 11th, 13th and so on. Those harmonic currents do three things inside the transformer: they add winding eddy-current loss, they add stray loss in the tank and structural steel, and they raise the hot-spot temperature on top of the 50 Hz temperature rise. A transformer rated on sine-wave current alone will therefore run hotter than nameplate conditions suggest.

Two standards frame the answer. IEC 61378-1 defines how converter transformers are rated: the rating is expressed in terms of the fundamental plus a declared harmonic spectrum, and the test regime includes harmonic loss measurement, so the rating is only meaningful when the spectrum is stated. IEEE C57.110 provides the derating method for standard transformers feeding non-sinusoidal loads, using the harmonic loss factor of the winding to work out how much the loading must be reduced to keep the hot spot within insulation limits. Contractually, the distortion itself is limited at the point of common coupling: IEEE 519 caps voltage THD at 5% for a 1–69 kV busbar with no single harmonic above 3%, and IEC 61000-3-6 provides the equivalent emission limits inside a medium-voltage installation.

Three practical rules follow. First, do not add capacitors to fix the power factor of a diode-rectifier load. A diode rectifier already has a high displacement factor, and adding an untuned capacitor bank at the same busbar creates a parallel resonance that can amplify one of the drive's own harmonics — the classic failure mode described in our transformer harmonics guide. Use a documented harmonic study with tuned or detuned filter banks, or a multi-pulse/AFE front end, instead. Second, specify special tests: a temperature-rise test with the actual harmonic spectrum, and harmonic loss measurement, are the only way to prove that the winding hot spot stays inside limits. Third, remember that filter banks and long cables change the voltage: a capacitor bank raises the fundamental voltage at the transformer terminals (over-excitation and extra core loss), and the capacitance of a long 33 kV cable can resonate with the filter, so the transformer's no-load flux density and its tap range may both need adjustment.

How do you check motor-starting voltage dip at the pump busbar?

The dip calculation is a three-line exercise with the MVA method, and it should be carried out for the worst combination of motors that the plant permits to start together. Using the example numbers below: a 2,000 kW, 6.6 kV motor with 96% efficiency and 0.90 power factor has a full-load current of about 202 A, and a direct-on-line start at six times full-load current draws about 13.9 MVA of starting power.

Motor FLC      = P / (√3 × U × η × cosφ)
               = 2000 / (1.732 × 6.6 × 0.96 × 0.90) ≈ 202 A
Start current  = 6 × 202 A ≈ 1,215 A
Start kVA      = √3 × 6.6 kV × 1,215 A ≈ 13.9 MVA

Busbar fault level (transformer 12.5 MVA, Z = 8.5%, upstream ignored)
               ≈ 12.5 / 0.085 ≈ 147 MVA

Voltage dip    ≈ Start kVA / Fault level = 13.9 / 147 ≈ 9.5%  (one motor)
Two motors together: 27.8 / 147 ≈ 18.9%   → exceeds typical limits

Two cautions are essential. The fault level must be calculated with IEC 60909 including the real upstream source impedance: ignoring the utility, as in the simplified line above, overstates the fault level and therefore understates the dip. And the acceptable limit is a plant specification, not a constant: values around 10–15% at the busbar and up to 15–20% at the motor terminals for a short transient are common industry practice, but the pump vendor's minimum voltage for re-acceleration and the membrane-protection logic may set a tighter limit.

If the dip is too large, the options in order of cost are: stagger the starts (free, but only if the process allows the delay); raise the transformer rating or lower its impedance (a cheaper transformer change than re-engineering the motors); fit soft starters or VFDs to the offending motors (which also fixes the starting current itself, since a VFD ramps at no more than 100% of full-load current); or add reactive support. Note that transformer energisation inrush — 5–12 times rated current for a few cycles — is a separate and much shorter event, covered in our inrush current guide, and must not be confused with motor starting when protection settings are chosen.

What is the six-step sizing method for a desalination plant transformer?

Step 1 — load list by process area. List every load with its rated power, its starting method and its process criticality. Desalination plants fail at this step more often than any other, because intake, pre-treatment and product-water pumping are forgotten.

Step 2 — demand and diversity. Not all loads run at maximum simultaneously. Apply a demand factor to the connected load and include a design spare for the plant's declared future expansion — additional RO trains are almost always part of the master plan.

Step 3 — convert to apparent power. Divide the diversified demand by the design power factor. For a VFD-dominated load the displacement factor is high (0.95–0.97 at full load), but the transformer must still be rated for the total apparent power including the harmonic content per IEC 61378-1.

Step 4 — check motor starting at the worst-case pump busbar, as shown above.

Step 5 — derate for harmonic duty. Apply the IEEE C57.110 harmonic loss factor for the declared spectrum, or specify to IEC 61378-1 with the spectrum stated, and choose an appropriate K-factor or converter-duty design — the choice procedure is in our K-factor guide.

Step 6 — derate for environment, then pick a standard rating. Hot coastal ambient (45–55 °C is routine on the Gulf coast) reduces cooling capability; a transformer rated for 50 °C ambient or an ONAF/ODAF design may be needed rather than the standard 40 °C ONAF unit. Apply the corrections, then round up to the next standard rating and re-check the impedance against the starting-dip result. Our derating guide covers the calculation method; a free implementation of the same arithmetic is in the engineering toolbox.

The worked example below applies all six steps to a nominal 100,000 m³/day SWRO plant.

The 12,000 kW figure for the RO high-pressure stage is not a guess: 100,000 m³/day at 45% recovery means about 9,260 m³/h of feed at roughly 62 bar, which is about 15.9 MW of hydraulic power; an isobaric energy-recovery device returns about 8.1 MW from the brine, leaving a net hydraulic demand near 7.8 MW, which becomes about 11.8–12 MW of electrical input once pump and motor efficiencies are applied. That is a specific energy of about 2.8 kWh/m³ for the RO stage — inside the range reported for modern plants and a useful sanity check on any load list.

Fire and oil policy should be settled at this stage too. An oil-immersed transformer standing outdoors on a coastal site is normal; the same unit inside an RO building or an electrical room is usually replaced by a dry-type transformer or an ester-filled unit to avoid a fire-rated oil containment room, a choice developed in our transformer oil types guide.

Why does an energy recovery device change the electrical design?

An energy-recovery device (ERD) takes the pressure energy in the reject brine and returns it to the feed, typically recovering 95–97% of the brine's hydraulic energy. It cuts the plant's biggest load by 40–50%, so it changes the transformer rating directly. It also changes where the remaining load sits, depending on the type:

• Isobaric (pressure-exchanger or piston) ERD: the brine directly pressurises feed water, so a separate booster or circulation pump is needed, often on the low-voltage or 6.6 kV bus with its own drive. The high-pressure pump then only makes up the differential.
• Turbine / turbocharger ERD: a turbine extracts the brine energy and assists the high-pressure pump on a common shaft, so the motor driving that shaft is smaller but the mechanical arrangement is more complex; electrically the load is smaller and more concentrated in one drive.

In both cases the electrical engineer must update the load list, re-run the harmonic study (the drives are now different sizes and counts) and re-check the starting or ramp duty on the busbar. A common specification error is to keep an old load list — and therefore an oversized transformer — after the ERD configuration changes from turbocharger to isobaric.

What does coastal corrosion demand from the transformer?

Marine atmosphere combines airborne chloride, high humidity, salt spray and intense ultraviolet radiation, and it attacks exactly the parts of a transformer that are hardest to repair: tank and radiator paint, bolted joints, breather assemblies and external hardware. The international framework for specifying the countermeasure is ISO 12944, with ISO 9223 for classifying the local atmosphere and ISO 1461 for galvanising. A coastal site is normally C5-M (marine); structures inside the splash or spray zone can reach CX, the extreme category.

The package matters more than any single line. A C5-M painted transformer with plated fasteners and a PVC-sheathed third-party cable will fail at the fastener and the cable long before the paint. Specification should therefore cover the whole installation, and should say explicitly that the coating system is verified by dry film thickness and adhesion tests before despatch — the pre-delivery verification logic in our pre-shipment inspection checklist. Coastal humidity is also why routine oil and gas checks matter more than inland: our dissolved gas analysis guide explains how moisture ingress shows up long before a fault.

How do you design N-1 redundancy and the switchgear arrangement?

For a 20 MVA firm load the three common arrangements are set out below. The choice is a commercial decision about the value of lost production, not a technical one, because all three are technically valid.

Three design rules make N-1 real rather than nominal. First, the switchgear must permit the transfer: duplicated busbars with a bus-tie breaker, or sectionalised single busbars, and a clearly documented switching sequence. Second, the load-shedding scheme must be tested, including the case where the surviving transformer also has to absorb the motors that restart. Third, the parallel-operation conditions must be met — same vector group, same ratio and tap position, impedance mismatch within limits, and correct phase sequence — because the plant will parallel the transformers during transfers. Those five conditions are set out in our parallel operation guide, and their relationship to the nameplate fields is explained in the nameplate reading guide.

Redundancy should extend beyond the transformer: a critical-spares agreement, or a fully spare unit sized to the largest rating in the plant, converts a transformer failure from a multi-month production outage into a crane-and-swap job. The lead-time arithmetic that makes this decision is in our delivery lead time guide, and it is one of the strongest arguments for buying from a manufacturer who stocks the standard ratings — see also how to audit a transformer manufacturer.

Which standards and specification fields belong on the transformer data sheet?

A desalination transformer data sheet should state, at minimum, the fields below. The right-hand column is the field's purpose; if a field cannot be justified, it is probably padding the document.

What tests and documentation prove fitness for a desalination project?

Routine tests to IEC 60076-1 — winding resistance, ratio and vector group, impedance and load loss, no-load loss and current, applied and induced voltage, insulation resistance — prove that the transformer is what the data sheet says it is. For desalination duty they are necessary but not sufficient, and three further items should be written into the purchase order:

• Harmonic loss measurement with the declared spectrum, so that the converter-duty rating can be verified rather than assumed.
• A temperature-rise test at the specified ambient and cooling stage, which for a hot coastal site may mean proving the unit at 50 °C ambient rather than 40 °C.
• A frequency-response analysis (FRA) fingerprint per IEC 60076-18, taken at the factory, so that any transport damage can be detected by comparison on site — the same logic as the transport precautions in our transportation and installation guide.

On site, the pre-energisation and commissioning sequence is the last opportunity to catch a mistake cheaply; the eight checks are listed in our commissioning tests checklist, and the fire and lightning exposure of a coastal substation is covered by the lightning protection guide and the bushing guide. Where noise matters — RO buildings near residential or environmental receptors — the limits and measurement method are in transformer noise level.

What are the eight most common specification mistakes?

1. Sizing on a rule of thumb. A kVA-per-m³/day figure hides differences in recovery, feed temperature, membrane age and ERD configuration, and produces transformers that are either oversized (money wasted) or undersized (dip and hot spots at commissioning).
2. Treating the VFD transformer as an ordinary distribution transformer. Without a stated harmonic spectrum and converter-duty design, the transformer is rated for a load it will never see.
3. Adding power-factor capacitors to a rectifier load. The result is resonance amplification, not correction.
4. Skipping the motor-starting study because "the biggest pumps have VFDs" — forgetting the DOL intake, backwash, transfer and outfall pumps on the same busbar.
5. Ignoring capacitor-bank over-excitation. Filter banks raise the fundamental voltage at the transformer terminals, so no-load flux and tap range must be checked for it.
6. Specifying an inland enclosure and paint system on a coast, then discovering the corrosion in year three.
7. Buying "two transformers" and calling it N-1 without a bus-tie, a transfer sequence or a tested load-shedding scheme.
8. Forgetting the spares and lead-time equation. Two installed units and no spares strategy means a single failure stops production for the whole manufacturing lead time.

Run the arithmetic rather than the folklore: put the load list, the demand and diversity factors, the starting-dip check and the ambient and harmonic derating into one calculation, and the transformer rating stops being a matter of opinion. The free QDTB engineering toolbox performs the same calculation automatically and produces the equipment list and budget from the same inputs.

Sources / 资料来源

• Source: IEC 60076 series — Power transformers, Parts 1 (General), 2 (Temperature rise), 3 (Insulation levels and dielectric tests), 5 (Ability to withstand short circuit), 7 (Loading guide for oil-immersed transformers), 10 (Determination of sound levels), 11 (Dry-type transformers) and 20 (Energy efficiency)
• Source: IEC 61378-1 — Converter transformers: transformers for industrial applications; and IEC 61378-3 — Application guide
• Source: IEC 60909 series — Short-circuit currents in three-phase a.c. systems
• Source: IEC 60076-18 — Measurement of frequency response (FRA)
• Source: IEC 60529 — Degrees of protection provided by enclosures (IP Code)
• Source: IEC 60815 series — Selection and dimensioning of high-voltage insulators intended for use in polluted conditions
• Source: IEC 61000-3-6 — Assessment of emission limits for distorting loads in MV and HV power systems; and IEC 61000-2-4 — Compatibility levels
• Source: IEC 60364 series — Low-voltage electrical installations; IEC 61936-1 — Power installations exceeding 1 kV a.c.; IEC 62271-200 — AC metal-enclosed switchgear and controlgear above 1 kV up to and including 52 kV; IEC 61439 series — Low-voltage switchgear and controlgear assemblies
• Source: IEEE Std 519 — Recommended Practice and Requirements for Harmonic Control in Electric Power Systems (voltage distortion limits at the point of common coupling)
• Source: IEEE Std C57.110 — Recommended Practice for Establishing Liquid-Immersed and Dry-Type Power Transformer Capability When Supplying Nonsinusoidal Load Currents
• Source: IEEE Std C57.91 — Guide for Loading Mineral-Oil-Immersed Transformers and Step-Voltage Regulators
• Source: ISO 12944 series — Corrosion protection of steel structures by protective paint systems, including corrosivity categories C5-M (marine) and CX
• Source: ISO 9223 — Corrosion of metals and alloys: corrosivity of atmospheres, classification, determination and estimation
• Source: ISO 1461 — Hot dip galvanized coatings on fabricated iron and steel articles
• Source: EU Commission Regulation (EU) No 548/2014 — Ecodesign requirements for small, medium and large power transformers
• Source: US Department of Energy 10 CFR Part 431 — Energy conservation standards for distribution transformers
• Source: GB/T 6451 — Technical parameters and requirements for three-phase oil-immersed power transformers; and GB 20052 — Minimum allowable values of energy efficiency and energy efficiency grades for power transformers
• Source: WHO Guidelines for Drinking-water Quality — health-based guidance for desalinated drinking water (total dissolved solids, boron)
• Source: EU Directive (EU) 2020/2184 — on the quality of water intended for human consumption
• Source: US EPA National Primary Drinking Water Regulations
• Source: International Desalination Association (IDA) — global desalination capacity and technology statistics
• Source: published SWRO plant performance studies — RO-loop specific energy around 2.0–3.0 kWh/m³ at best-in-class plants, including a verified 1.794 kWh/m³ at 40% recovery on a 2,500 m³/day plant
• Source: QDTB Engineering Team — industry experience / 行业经验值 for desalination substation layering and typical rating bands, demand, diversity and design-spare factors, motor-starting voltage-dip limits, ERD electrical configuration practice, harmonic-duty specification practice, coating and enclosure practice for coastal sites, cost-driver ranking and critical-spares lead time