Causes of Unbalanced Current in Transformer Differential Protection — and How to Eliminate It
Why differential protection needs a margin it can live with
Differential protection is the main protection of a power transformer. Transformers rated 6300 kVA and above are normally fitted with longitudinal (biased) differential protection, which works on the circulating-current principle: current transformers are installed on both sides of the transformer, and their secondaries are wired so that the two secondary currents circulate through the relay circuit. The relay operating coil is connected across the secondary terminals.
During normal operation, and for any fault outside the protected zone, the two secondary currents are equal in magnitude and opposite in direction. Their difference — the current through the relay — is zero, so the relay does not operate. That is the theory. In practice the difference is never exactly zero, because a number of physical effects add a residual current to the differential branch. This residual is called the unbalanced current; in English-language literature you will also meet it as spill current or false differential current.
The protection does not have to make the unbalanced current disappear — it has to ride through it. Time grading alone cannot do that: the relay must be fast enough to trip on an internal fault within a few cycles, while the transient part of the unbalanced current lasts just as long. So the causes have to be understood one by one, and each one handled with a matching countermeasure.
Key takeaways
There are five classical sources of unbalanced current: magnetizing inrush, the 30° phase displacement between the two windings, the gap between calculated and actual CT ratio, dissimilar CT types at the two ends, and on-load tap changing.
Only two of them — phase displacement and ratio mismatch — can be reduced by choosing CT connections and CT ratios. The other three are inherent and must be covered by the relay characteristic or by the setting.
Inrush is handled by second-harmonic restraint, waveform (dead-angle) blocking or a quick-saturating current transformer in the differential circuit, not by raising the pick-up value alone.
Whatever is left over after compensation is called residual unbalanced current, and it must be added to the differential pick-up setting; a smaller residual means a more sensitive protection.
The five causes of unbalanced current
In order of appearance in the differential circuit:
Magnetizing inrush current when the transformer is energised
Different phase angles of the currents at the two ends
Difference between the calculated transformation ratio and the actual ratio
Different types of current transformer at the two ends
Tap changing on the on-load tap changer (OLTC)

1. Unbalanced current caused by magnetizing inrush
Inrush current flows through one side of the transformer only, so it cannot be balanced out through the current transformers — the differential circuit sees all of it. In steady state the excitation current is small, a few per cent of rated current at most, and its contribution is negligible. The problem appears at switching: when the transformer is energised, or when the supply is restored after an external fault has been cleared, the core can be driven deep into saturation and the excitation current rises to 5–10 times rated current.
That inrush is not a sine wave. It is a spiky, heavily offset wave rich in d.c. (aperiodic) and harmonic components, with the second harmonic dominating. Its magnitude and rate of decay depend on the point on the voltage wave at which the breaker closes, the remanent flux in the core, the source impedance, the transformer rating and the core material. If the breaker happens to close at the voltage peak there is almost no inrush — only normal excitation current. In a three-phase transformer, with the phases closing at different instants, at least two phases will always show some inrush.
How it is handled
Second-harmonic restraint. Inrush contains a large share of second harmonic; an internal fault current is essentially fundamental. The relay blocks — or biases back — whenever the second-harmonic content of the differential current exceeds a threshold, typically 15–20 %. This is the standard method in numerical relays and the first thing to check when a transformer trips on energisation.
Waveform (deadangle) discrimination. The rectified inrush waveform has intervals where the current is almost zero, which a genuine fault current does not have. When those intervals persist beyond a set time, the relay classifies the event as inrush and blocks.
Quick-saturating current transformer in the differential circuit. The classic BCH-type relay uses a saturating converter between the CTs and the operating coil. Its core saturates quickly under a large current containing a d.c. component, so the transient unbalanced current is largely prevented from reaching the relay coil and the protection stays stable on external faults. The decay of the aperiodic component is fast — generally complete within about two cycles — after which the full short-circuit current is transferred and the relay operates sensitively on an internal fault.
In numerical relays all three functions are software blocks; only their settings need to be coordinated with the transformer rating and with the CT saturation level.
2. Unbalanced current caused by the 30° phase displacement
Power transformers are usually connected Yd11, so the currents on the two sides differ in phase by 30°. Even with perfectly matched magnitudes, the vector difference appears in the differential circuit as an unbalanced current of the order of the load current itself.
The classical remedy is phase compensation by connection: the three CTs on the star side are connected in delta and the three CTs on the delta side in star, which rotates the secondary currents back into phase. The penalty is a factor of √3: after compensation the current in the differential branch on the delta side is 1.732 times the CT secondary current, so the ratio of the CTs on that side must be raised by 1.732 to bring the two branches back into balance. In numerical relays the same result is obtained by entering a phase-compensation matrix together with the amplitude balance coefficients, and no physical delta connection is needed — which also avoids the risk of an open delta circuit.
3. Unbalanced current caused by calculated versus actual CT ratio
CTs come in standard ratios, and the transformer ratio is fixed by the design; rarely do the two line up exactly. The mismatch shows up directly as current in the differential branch.
In practice the CTs are selected as close as possible to the transformer ratio once the design is frozen. Take a 31.5 MVA, 110/35 kV transformer with Yd11 connection, CTs of 300/5 A on the high-voltage side and 600/5 A on the low-voltage side:
Rated current, HV side: 31 500 / (√3 × 110) = 165.3 A → CT secondary 165.3 × 5/300 = 2.76 A
Rated current, LV side: 31 500 / (√3 × 35) = 519.6 A → CT secondary 519.6 × 5/600 = 4.33 A
After delta compensation of the HV side: 2.76 × 1.732 = 4.77 A
Unbalanced current in the differential branch: 4.77 − 4.33 = 0.44 A
The classical countermeasure is the balance (restraint) winding of a quick-saturating relay, connected on the branch with the smaller current. Because the balance winding and the operating winding are wound on the same limb, the ampere-turns of the balance winding can be made to cancel the ampere-turns produced by the ratio mismatch, leaving no flux and therefore no operation. In practice the balance can only be set in whole turns, so a residual unbalanced current always remains and must be included in the setting calculation. In numerical relays the same job is done by the balance coefficients; the residual is usually much smaller, but it is still there and should not be ignored.
4. Unbalanced current caused by different CT types at the two ends
If the CTs at the two ends have different construction — different cores, different accuracy limits, different saturation behaviour — their transient response during an external fault will not match, and the difference appears as a comparatively large unbalanced current precisely when the through-current is largest.
Two measures apply: use CTs of the same type, class and rating at both ends wherever the design allows, and raise the differential pick-up current to cover the residual mismatch. Where the two ends physically cannot use the same CT type, that mismatch has to be documented in the setting calculation rather than assumed away.
5. Unbalanced current caused by on-load tap changing
The OLTC is the normal means of voltage control, and changing tap changes the transformer ratio. If the differential protection has already been balanced for one tap position — by CT connection or by balance coefficients — every subsequent tap step introduces a new unbalanced current that no amount of balancing can remove, because the operating condition itself has moved.
The accepted approach is to allow for it in the setting: the tap range is quantified, the worst-case ratio deviation is calculated, and the differential pick-up current is raised accordingly. Numerical relays with tap-position input can compensate automatically over the full range, which is a good reason to specify that input when an OLTC is present.
Summary: which causes can be eliminated and which cannot
|
Cause |
Can it be eliminated? |
Countermeasure |
|
Magnetizing inrush |
No — inherent to energisation |
Second-harmonic restraint, waveform blocking, quick-saturating CT |
|
30° phase displacement |
Yes — largely |
Delta/star CT connection or software phase compensation, ratio ×1.732 |
|
Calculated vs. actual CT ratio |
Partly — residual remains |
Balance winding / balance coefficients, setting margin |
|
Different CT types |
No — but reducible |
Same CT type at both ends, higher pick-up |
|
OLTC operation |
No — inherent to tap changing |
Include tap range in the setting, use tap-position compensation |
Conclusion
Unbalanced current is not a defect of the transformer; it is the sum of five physically different effects that happen to appear in the same measuring branch. Two of them — the 30° phase displacement and the gap between calculated and actual CT ratio — can be brought close to zero by choosing the CT connection and the CT ratio properly, and by trimming the balance. The remaining three — magnetizing inrush, dissimilar CT types and on-load tap changing — cannot be eliminated at all, and must be absorbed by the relay characteristic and by the setting.
The practical rule that follows is simple: minimise what can be minimised, quantify what cannot, and add the residual to the pick-up value. The smaller the unbalanced current left in the differential branch, the more sensitive the protection can be set — and the more reliably the transformer is protected, without paying for that sensitivity with nuisance tripping on energisation.
FAQ
What is unbalanced current in transformer differential protection?
It is the current that flows in the differential branch even when there is no internal fault. It comes from magnetizing inrush, the 30° phase shift between the windings, mismatch between calculated and actual CT ratios, dissimilar CT types, and tap changes. It is also called spill current or false differential current.
Why does a transformer trip when it is switched on?
Almost always because of magnetizing inrush. Inrush current flows on one side only and reaches 5–10 times rated current, with a large second-harmonic content. Either the second-harmonic restraint setting is too low or too high for the actual core and CT saturation level, or the blocking function has been disabled during commissioning.
Can unbalanced current be removed completely?
No. Phase displacement and ratio mismatch can be reduced to a small residual; inrush, dissimilar CT types and tap changing are inherent. The protection is always set with a margin that covers the residual.
Which standard governs transformer differential protection?
In China, GB/T 14285 (relaying protection and automatic devices) and DL/T 584 give the application and setting requirements; IEC 60255 covers the relay characteristics and testing. CT requirements for differential protection are defined by the CT accuracy-limit factor rather than by the metering class.
If a case on your project does not fit the pattern above — unusual vector group, wide tap range, very different CT types at the two ends — send us the nameplate and the CT data and we will go through the setting with you.
Telephone / WeChat: +86 158 5325 2696 E-mail: hfy@hengfengyou.com











