How-to
Estimating transformer inrush current for protection coordination
Short answer
The rules of thumb are screening gates, not coordination inputs: reported inrush runs 8 to 30 times full-load current and lasts 10 cycles to a minute. Plot the decay curve Ipu = 6 × t^-0.31 with t in seconds — it passes through 25× at the first peak, 12× five cycles in and 3× at 10 s — keep the protective device curve above it and below the transformer damage curve, and handle cold-load pickup separately.
On this page
The 12× full-load current gate is an adequate screening gate and an inadequate coordination input, and the other two numbers in circulation are not independent estimates: 25×, 12× and 3× FLC are three points on one decay curve, Ipu = 6 × t^-0.31, with t in seconds and Ipu in per-unit of full-load current. Plot that curve, keep the fuse or relay curve above it, and keep the same device curve below the transformer's through-fault damage curve. Magnetizing inrush and cold-load pickup are different events and get different settings. Reported inrush spans 8 to 30 times full-load current and lasts from 10 cycles on small units to a minute on large ones, which is exactly why one multiplier cannot place a point on a time-current plot.
What the screening points are for #
A single-point criterion answers one question: will this device survive energization? The 12× FLC figure is the screening point quoted at 0.1 s for transformers above 3 MVA self-cooled rating, and many engineers apply it at any size; below 3 MVA the common screening value is 8× FLC at the same instant, because smaller cores decay faster. These are pass/fail gates for a fuse or an instantaneous element, not predictions of what your bank will draw.
| Screening point | Stated basis | What it decides |
|---|---|---|
| 25× FLC | First-peak check | Instantaneous element pickup |
| 12× FLC | Single-point criterion for units above 3 MVA self-cooled | Fuse minimum melting and short-delay elements |
| 8× FLC | Same criterion applied below 3 MVA | Same check on a faster-decay assumption |
| 3× FLC at 10 s | Rule of thumb quoted without a stated size boundary | Long-time elements; a point on the curve below |
Why no single multiplier fits #
Inrush is worst when the breaker closes near a voltage zero: the inductive circuit then needs a flux change of twice the normal peak over the next half cycle, and residual flux left in the core from the previous shutdown adds to or subtracts from it. Power transformers normally run near the knee of the saturation curve, so that extra flux saturates the iron and the magnetizing current spikes.
Measured energization currents run from 8 to 30 times full-load current, with durations from 10 cycles on small units to a minute on large ones; individual references quote narrower bands (2-8×, 3-14×, 10-15×) for the size classes they studied, which tells you the spread is real, not sloppy measurement. In an IEEE survey of relaying practice, inrush was named the root cause of false tripping by 75 percent of the respondents who reported trouble — and the respondents with no trouble used time delays of 1/3 to 1/2 s, reduced sensitivity, or both. That is the whole menu: time, or sensitivity.
The curve to plot instead #
Use Ipu = 6 × t^-0.31, where t is the time in seconds after energization and Ipu is in per-unit of full-load current. That t is in seconds is confirmed by arithmetic, not assumption: 6 × 0.01^-0.31 = 25.0 pu at the first peak, 6 × 0.1^-0.31 = 12.3 pu five cycles in, and 6 × 10^-0.31 = 2.9 pu reproduce the three rules of thumb as points on this one curve. On a 50 Hz system the first peak lands at 0.01 s, one half cycle; at 60 Hz it lands at 0.0083 s, where the curve reads 26.5 pu.
The companion tail expression sometimes quoted, Ipu = 5 × t^-0.6 - 1.8 for t above 5 s, does not work as an envelope. At 5 s the main branch gives 3.6 pu and the tail expression gives 0.1 pu — a decay curve cannot drop that far between adjacent points — and the tail reaches zero near 5.5 s, so it cannot represent the minute-long tails reported for large units. Use the main branch across the whole time axis of the study: at 60 s it still reads 1.7 pu, consistent with the reported durations.
The per-unit value refers to either winding; multiply by the full-load current of the winding the device watches to get amps.
How to run the coordination check #
- Full-load current from the rating: FLC = S / (√3 × V), with S in volt-amperes and V the line-to-line voltage of the winding the device protects. The result is an estimate; the printed nameplate current wins if it differs.
- Tabulate the envelope at the first peak, five cycles at 50 Hz, 0.5, 1 and 10 s, plus 60 s for large banks, on the side the device watches. Magnetizing inrush is primary-side current, so a secondary-side device does not see it at all.
- Plot envelope and device curves on the same log-log time-current axes: fuse minimum-melting curves, not average-melting, and relay curves including the instantaneous element.
- The device curve sits above the envelope at every point, with the margin your coordination practice carries. The first-peak and five-cycle points are what decide fuses and instantaneous elements.
- The device must still clear faults inside the transformer's through-fault damage curve. The fault current the bank delivers is a separate calculation from the inrush it draws — see closed delta transformer fault current calculation — and the thermal limit under sustained load is its own check, covered under dry-type transformer hot spot temperature rise.
- Where a sensitive element cannot be raised above the envelope, buy ride-through with time: the practice that worked in the survey was 1/3 to 1/2 s of delay, reduced sensitivity, or both.
- After an inrush trip, pull the relay event record or fuse evidence before touching settings; the trace shows which point of the envelope was crossed. Capturing a new trace means energized switching: NFPA 70E, 2024 edition, the United States workplace electrical safety standard, sets the approach-boundary and PPE requirements — confirm them against your site program, the authority having jurisdiction and the equipment documentation, and never open an energized panel or terminal box; de-energize and verify absence of voltage first.
Magnetizing inrush is not cold-load pickup #
Magnetizing inrush happens on every energization of a dead transformer, whether its load is connected or not. Cold-load pickup happens after an outage: the load returns with its diversity lost and its motors restarting together, so current above full load persists far longer than any magnetizing transient. Some relay literature calls the first event hot load and the second cold load, which confuses people, because the magnetizing event occurs whether the bank is hot, cold or unloaded. Coordinate short-time elements against the inrush envelope and long-time elements against cold-load pickup built from the plant's actual load mix; no single multiplier covers both.
Worked example: inrush envelope for a 1500 kVA, 11 kV to 433 V transformer #
Inputs: a 1500 kVA bank with an 11 kV primary, as used on 11 kV site distribution; the 433 V secondary nominal marks a 50 Hz system, so the first peak is the half-cycle point defined above. The envelope is Ipu = 6 × t^-0.31 with t in seconds. The below-3 MVA screening gate is 8× FLC at the same instant.
Rule in words: full-load current is the three-phase rating divided by root three times the line voltage, and each envelope point is that current times the per-unit value at that time.
Arithmetic:
FLC = 1,500,000 VA / (1.732 × 11,000 V) = 1,500,000 / 19,052 = 78.7 A
| Time after energization | Envelope, per-unit | Envelope at 11 kV |
|---|---|---|
| First peak (6 × 4.17 = 25.0 pu) | 25.0 | 1970 A |
| Five cycles (6 × 2.04 = 12.3 pu) | 12.3 | 965 A |
| 0.5 s (6 × 1.24) | 7.4 | 586 A |
| 1 s | 6.0 | 472 A |
| 10 s (6 × 0.49) | 2.9 | 231 A |
Decisions for this bank: the fuse or relay curve on the 11 kV side clears 1970 A at the first peak and 965 A five cycles in, or it does not coordinate — those two points are the pass/fail line. A fuse chosen against the small-transformer gate alone (8 × 78.7 A = 630 A at that instant) sits 35 percent below the envelope and can blow on a bad closing, so coordinate to the envelope. Set any instantaneous element above 1970 A or delay it past the first cycle, and confirm the inverse-time element does not operate at 965 A at that point. On a 60 Hz system the first peak lands at 0.0083 s, the envelope reads 26.5 pu = 2090 A, and that is the number to clear instead of 1970 A.
Questions asked
Can a 4225 kVA, 11 kV generator energize all fourteen site transformers at once?
No. At five cycles on a 50 Hz system the envelope is 12.3 × 52.5 A = 644 A per 1000 kVA bank and 12.3 × 78.7 A = 965 A per 1500 kVA bank; eight of one and six of the other is about 10,900 A against a machine rated 222 A at 11 kV. The generator's subtransient reactance holds the actual current far below that and turns the difference into voltage dip, so energize the banks one at a time with the secondaries open.
Does synchronizing onto an already energized transformer cause inrush?
No magnetizing inrush: closing a source onto a bank that is already energized at matching voltage is not the dead-transformer event. The case to watch is the parallel bank beside it — energizing one transformer can disturb the flux of an already-energized neighbor and stretch out its inrush, a documented effect in parallel-bank substations — so set the bus protection to ride through both banks, not one.
Why does the same transformer trip protection only on some energizations?
The peak depends on the closing angle and on the residual flux left in the core by the previous shutdown. Closing near a voltage zero asks for a flux swing of twice the normal peak, and residual flux of the same polarity adds to it and drives the core deeper into saturation. Same bank, different day: anywhere across the reported 8 to 30 times full-load current.
When is the plotted envelope not enough?
Three cases justify manufacturer test data or a transient simulation: a generator or other weak source energizing the bank, because source impedance changes both magnitude and decay; parallel transformer banks on one bus; and protection that must stay sensitive to internal faults while riding through inrush. For fuse and inverse-time relay coordination on a stiff supply, the envelope is the working tool.