A distribution transformer is bought once and then sits energised for thirty years, paying its losses every hour and getting a little older every time it runs hot. Sizing it is not a matter of adding up nameplates: it is a judgement about how much of the connected load will actually be on at the same time, how much will be added later, and how hard the unit can be run in a Punjab summer. And one number on the nameplate that has nothing to do with capacity, the percent impedance, decides what the switchgear underneath it has to be able to interrupt.
From connected load to demand
Connected load is the sum of everything wired to the transformer. Maximum demand is the highest load that actually occurs, averaged over a short window. The ratio between them is the demand factor, and it is always less than one because motors are not all running, heaters cycle, lights are off in empty rooms and no factory runs every machine at full load at the same minute. A group of loads also has diversity between its parts: the peak of the compressor room and the peak of the offices do not fall at the same hour, so the site's peak is less than the sum of the group peaks.
On an existing site there is no reason to guess. A logger on the incomer for at least a week, at 15-minute averages and ideally through the hottest weeks of the year when air conditioning peaks, gives the maximum demand directly. The MDI figure on a commercial or industrial bill is the same thing measured by the utility: the highest 30-minute average demand in the month, and a year of bills shows how it moves with the seasons. For a new site there is nothing to measure, so demand factors come from experience and from published tables: lighting close to one, general-purpose sockets much lower, motors according to the process. Treat any table value as a rule of thumb, and be honest about which loads genuinely coincide. A demand factor of 70 % on a mixed industrial site, the figure in the worked example, is a plausible starting point, not a standard.
kVA, power factor and growth
A transformer is rated in kVA, not kW, because what heats the windings is current, and current depends on apparent power. Real power divided by the power factor gives the apparent power the transformer must carry: 175 kW at a power factor of 0.85 is 205.9 kVA, but the same 175 kW at 0.95 would be 184 kVA. Correcting the power factor at the LV board, as the power factor guide explains, frees transformer capacity in the same way it frees cable and switchgear capacity, and is often the cheapest way to defer the next size.
Growth is the next allowance. Twenty percent is a common planning figure for a site that expects to add machines; for a site that is already full, use less and spend the money elsewhere. Note that the standard ratings themselves add margin: the IEC 60076-1 preferred series steps up by about 25 % each time (250, 315, 400, 500, 630 kVA), so rounding up to the next size usually adds something on top of what you asked for. The transformer sizing calculator carries all of this through and reports the loading the unit will actually see.
Why not load it to 100 %
The nameplate rating is what the transformer can carry continuously at the reference conditions of IEC 60076-1: an ambient of no more than 40 °C, a monthly average of 30 °C in the hottest month, and a yearly average of 20 °C. At that load and that ambient the design puts the hottest point in the windings, the hot-spot, at about 98 °C, and the standard's thermal model treats that as the temperature at which insulation ages at its normal rate.
Insulation ageing is a chemical reaction and follows the Arrhenius relation, which lighting engineers and transformer engineers alike simplify to a rule of thumb known as the six-degree rule, after Montsinger: the rate of ageing roughly doubles for every 6 °C the hot-spot rises above its reference. IEC 60076-7 writes it as a relative ageing rate of 2(θh − 98) ÷ 6 for ordinary kraft paper. A unit fully loaded in a substation whose yearly average is 10 °C above the standard's 20 °C runs its hot-spot near 108 °C and ages at about 210/6 = 3.2 times the normal rate; a thirty-year design life becomes ten. Karachi's summer and a Lahore substation with dust-choked radiator fins are exactly that case.
Loading to 80 % is the practical answer. Load losses go with the square of the current, so at 80 % they are 64 % of the full-load value; the winding temperature rise falls accordingly and the hot-spot comes back down toward the reference even in a hot room. The 20 % in hand also absorbs the growth that was guessed rather than measured. IEC 60076-7 does permit loading above the nameplate for limited periods, trading life for capacity through emergencies, but that is a tool for the operator, not a basis for the design.
Percent impedance and the fault current
The percent impedance, Z %, is measured in the factory: the secondary terminals are short-circuited and the primary voltage raised from zero until rated current flows. The voltage needed, as a percentage of the rated voltage, is Z %; the power drawn during the test is the load loss. A 4.5 % transformer drives rated current into a dead short with only 4.5 % of its rated voltage, so at full voltage a short at its terminals draws about 1 ÷ 0.045 = 22 times rated current, limited only by its own impedance:
This is the infinite-bus shortcut. It assumes the 11 kV system upstream is stiff enough that the transformer's own impedance is all that matters, which at distribution level is nearly true: a 250 MVA utility fault level looks like 0.126 % on a 315 kVA base, against the 4.5 % of the transformer, and takes only 3 % off the result. The shortcut is therefore the conservative figure, and the right one to use when the utility cannot tell you its fault level.
Why it matters: every breaker, fuse and busbar on the LV side must be able to interrupt and withstand that current. A circuit breaker's rated breaking capacity, Icu in IEC 60947-2 terms, has to exceed the prospective fault current at its terminals, and miniature circuit breakers are commonly rated at only 6 or 10 kA. Below a 315 kVA transformer, where the terminal fault current is above 10 kA, the main board needs moulded-case or air circuit breakers, or back-up protection that the manufacturer has tested. Further down the installation the cable impedance reduces the fault current, and a smaller rating becomes acceptable; IEC 60909-0 is the method for working that out, with a voltage factor c of 1.05 or 1.10 at low voltage and a correction factor on the transformer impedance that both raise the figure above the shortcut, and with the contribution of running motors added on top.
Z % also sets the voltage regulation. At rated current the terminal voltage drops by roughly R % × cos φ + X % × sin φ, a few percent for a 4.5 % unit at a lagging power factor, which is why makers set the no-load secondary at 415 or 433 V rather than 400 V. Small distribution transformers are built around 4 % to keep that regulation tight; above 630 kVA the fault current would become unmanageable, so 6 % and higher is used and the regulation is accepted. IEC 60076-1 allows a tolerance of ±10 % on the guaranteed impedance, so a nominal 4.5 % may arrive as 4.05 %, and the switchgear should be rated for the lower figure.
Losses: no-load and load
A transformer has two kinds of loss and they behave differently. The no-load loss is in the core, from hysteresis and eddy currents driven by the magnetising flux, and it is there whenever the unit is energised, whether it is carrying full load or nothing at all: 8 760 hours a year. If the no-load loss is 400 W, that is 3 500 kWh a year on the bill before a single machine is switched on. The load loss is in the windings, the I²R loss plus stray losses, and goes with the square of the current: 64 % of its full-load value at 80 % load, 25 % at half load, nothing at night.
Efficiency is highest where the two losses are equal. With load losses around ten times the no-load losses, as is common for a distribution design, that point is near √(1 ÷ 10), about 30 % of rating, which is one reason a lightly loaded transformer is not an efficient one and a heavily oversized one is not either. A tender should ask for both figures and evaluate them with capitalised loss factors, a price per watt of no-load loss and a lower price per watt of load loss, so that a unit with a better core wins over one that is merely cheaper. IEC 60076-20 defines the efficiency indices and IEC 60076-1 sets the tolerances on guaranteed losses.
The example in numbers
A factory in Faisalabad: 250 kW connected, 70 % demand factor, power factor 0.85, 20 % growth, planned loading 80 %, an 11 kV/400 V transformer with 4.5 % impedance and a 250 MVA utility fault level:
- Demand: 250 × 0.70 = 175.0 kW; 175 ÷ 0.85 = 205.9 kVA. Growth: 247.1 kVA. Planned loading: 247.1 ÷ 0.8 = 308.8 kVA, so a 315 kVA unit.
- Full-load current: 454.7 A at 400 V and 16.5 A at 11 kV.
- Fault current, infinite bus: 454.7 ÷ 0.045 = 10 104 A, about 10.1 kA. With the 250 MVA source: 9 828 A, about 9.8 kA, a fault level of 6.81 MVA.
- Loading: 65.4 % today, 78.4 % with the growth, within the planned 80 %.
The LV main board is specified for a prospective fault current above 10 kA, in practice 16 kA or 25 kA switchgear, and the 11 kV protection is sized for 16.5 A of full-load current with the transformer's inrush in mind. The 454.7 A on the LV side is the starting point for the busbar and the incomer cable, which the power and current calculator and the cable sizing tool take from here.
Choosing between one large and two smaller units
For the example, two 160 kVA units in parallel would give 320 kVA against the single 315 kVA, and two 200 kVA units would give 400 kVA. The single unit is cheaper, needs one HV cell and one LV incomer, and has one set of no-load losses; for most sites up to about a megawatt it is the right choice, with space and an HV cell reserved for a second unit if the load grows.
Two units earn their cost where the load matters more than the money: a process that cannot stop, a hospital, a data room. Even then, be clear about what the second unit buys. Two paralleled 200 kVA units carry the full load with one out of service only if the load is below 200 kVA, and their combined fault current is nearly double that of one, so the switchgear rating rises. Paralleling also has conditions: the same voltage ratio and vector group, and percent impedances close to each other (within about 10 % is the usual requirement) so that the load divides in proportion to the ratings rather than piling onto the lower-impedance unit. A common compromise is two units on two LV boards joined by a normally open bus coupler: no paralleling in normal service, full flexibility in a fault, and each board rated only for its own transformer. That arrangement is also where a standby generator connects most naturally, one board at a time.
References
- IEC 60076-1:2011, Power transformers — Part 1: General — preferred ratings, service conditions, tolerances on impedance and losses
- IEC 60076-7:2018, Power transformers — Part 7: Loading guide for mineral-oil-immersed power transformers — hot-spot temperature, relative ageing rate, loading beyond nameplate
- IEC 60909-0:2016, Short-circuit currents in three-phase AC systems — Part 0: Calculation of currents
- IEC 60947-2, Low-voltage switchgear and controlgear — Part 2: Circuit-breakers — rated short-circuit breaking capacities Icu and Ics
- IEEE Std C57.91-2011, IEEE Guide for Loading Mineral-Oil-Immersed Transformers and Step-Voltage Regulators — the North American loading guide; same ageing physics