What goes in must come out, after you account for the transformer
A differential relay applies Kirchhoff's current law to a protected zone. Measure every current entering and leaving the zone; if they don't sum to zero, current is going somewhere it shouldn't, and that is a fault inside the zone. For a bus or a line, that is nearly the whole story.
A transformer is harder because, even with nothing wrong, the currents on its two sides do not simply cancel:
- Different magnitudes. The 34.5 kV side carries several times the current of the HV side, measured through CTs of different ratios.
- Phase shift. A YNd1 MPT shifts the LV currents by 30°.
- Zero sequence. A grounded-wye winding can carry I₀ that the delta side never sees.
- Magnetising current flows only on the energised side: small in steady state, enormous during inrush or overexcitation.
- On-load tap changers change the ratio in service, so the "balance" you set on day one does not hold at every tap.
- CT errors, small in steady state and large if a CT saturates (Module 2.3).
Every 87T setting exists to manage one of those effects. This module walks through them in order, with the operate–restraint plane as the central picture.
TAP scaling: putting both windings on one per-unit base
A numerical relay first scales each winding's secondary current by a TAP so that rated power on a chosen MVA base reads 1.0 pu on every winding:
It also applies internal phase-angle compensation for the vector group and removes zero-sequence current from any grounded-wye winding, so an external ground fault on the HV system does not look like differential current. That is why modern practice connects every CT in wye and lets the relay compensate, rather than delta-connecting CTs as electromechanical schemes did.
Worked example
A representative collector MPT is rated 100/133/167 MVA (ONAN/ONAF/ONAF), 230/34.5 kV YNd1. It has 1200:5 multi-ratio HV bushing CTs used at full ratio and 3000:5 LV CTs. Compute the TAPs on a 100 MVA base and a 167 MVA base. What does O87P = 0.30 pu mean in primary MVA on each?
| Base | TAP 1 | TAP 2 | O87P 0.30 pu = … |
|---|---|---|---|
| 100 MVA (chosen base) | 1.05 A | 2.79 A | 30.0 MVA primary |
| 167 MVA (top rating) | 1.75 A | 4.66 A | 50.1 MVA primary |
The dual-slope characteristic
The relay computes an operate quantity and a restraint quantity for each phase. This lab uses the average-magnitude definition (others use maximum or sum; check your relay):
Here and are the TAP-scaled, compensated winding currents, not the positive- and negative-sequence currents used elsewhere in the course.
A through-current with perfect CTs gives . An internal fault fed from one side gives . The characteristic sits between the two:
| Setting | Typical range | Representative | What it guards against |
|---|---|---|---|
| O87P minimum pickup | 0.15–0.40 pu | 0.20–0.30 pu | Steady-state excitation current and small errors at light load |
| SLP1 | 15–40% | 25–35% | Proportional errors: OLTC range plus CT ratio error |
| IRS1 breakpoint | 2–4 pu | 3 pu | Where CT saturation becomes credible |
| SLP2 | 50–100% | 60–85% | Asymmetric CT saturation on heavy through-faults |
| 87U unrestrained | 8–12 pu | 10 pu | Severe internal faults, cleared without harmonic delay |
| 2nd-harmonic block | 10–25% | 15% | Energisation and sympathetic inrush |
| 5th-harmonic block | 25–45% | 35% | Overexcitation (with 24 as the primary protection) |
Sizing SLP1 from the tap changer
Suppose the same 230 kV MPT has an HV on-load tap changer spanning 195.5–253 kV (a representative range; many MPTs have a de-energised tap changer instead, with a much smaller in-service error). Relative to the nominal ratio used for TAP:
Harmonics: telling inrush and overexcitation from faults
Inrush. Energising a transformer at an unfavourable point on the voltage wave, especially with remanent core flux, drives the core into saturation for part of each cycle. IEEE C37.91 and manufacturer literature put the instantaneous peak at roughly 8–30× rated current depending on size and design (12–15× is commonly cited for large units; treat these as representative), decaying over hundreds of milliseconds to seconds. It flows on one side only, so it is all differential current. The waveform is asymmetric and rich in 2nd harmonic, which is the basis for 2nd-harmonic blocking or restraint.
Overexcitation. High voltage or low frequency (V/Hz above about 1.05–1.10 pu) saturates the core symmetrically. The exciting current carries odd harmonics; the 3rd is trapped in delta windings and removed by zero-sequence filtering, so the 5th harmonic is the usable signature.
Restricted earth fault (87N): the faults 87T cannot see
For a ground fault at a fraction of the way from the neutral of a grounded-wye winding, the driving voltage is only of phase voltage, and only of the turns couple the fault current to the line side. The phase current change scales roughly as , and 87T's zero-sequence removal subtracts a further third. Faults in the bottom 20–30% of the winding can be invisible to 87T.
REF compares the neutral CT current with the residual sum of the phase CTs () for that winding only. It is supervised directionally and checked for CT saturation, and it sees a current proportional to . REF matters most on impedance-grounded windings where fault current is limited.
Simplified, resistance-grounded-style approximation for teaching. On a solidly grounded winding the fault current near the neutral stays higher because winding leakage reactance falls too, but the phase-side reflection still shrinks with x. Both curves are on a common per-unit base for comparison.
Overexcitation (24) and the rest of the MPT suite
The 5th-harmonic block keeps 87T secure during overexcitation, but something still has to protect the core. That is the Volts-per-Hertz element (24), usually a definite-time alarm stage plus an inverse-time trip stage, coordinated below the manufacturer's overexcitation capability curve.
Open question: loaded vs unloaded. The pickup check above uses the 1.05 pu full-load limit. At a collector the MPT often sits energised overnight with no load, while cable charging lifts the 34.5 kV voltage, and C57.12.00 allows up to 1.10 pu continuously at no load. A pickup below 1.10 pu (the lab default is 1.06 pu) can therefore time out on a healthy, unloaded MPT: watch the 1.08 pu readout. Practices differ: some set the inverse-time pickup against the no-load limit, others keep it lower and rely on a definite-time alarm stage (not modelled here) plus operating procedures. Agree the basis with the transformer manufacturer and your utility.
Withstand curve is illustrative only. Always coordinate against the manufacturer's overexcitation capability curve, and against the generator's where the MPT is a GSU.
The complete MPT package in a collector substation typically includes: dual-slope 87T with 87U, REF on grounded-wye windings, 24, sudden pressure (63) and other mechanical devices (gas accumulation, oil and winding temperature), backup phase and ground overcurrent (50/51, 51G on the neutral), and HV breaker-failure logic that keys direct transfer trip to the remote utility terminal.
- TAP = MVA·1000/(√3·kV·CTR). Choose one MVA base and keep every per-unit setting on it.
- SLP1 covers proportional errors: tap range plus CT ratio error. Size it on the actual restraint definition, not just the sum of percentages.
- SLP2 (or an external-fault detector) provides security against CT saturation on heavy through-faults.
- 2nd harmonic indicates inrush; 5th indicates overexcitation. Cross-blocking protects against low-H2 phases.
- 87U sits above the inrush fundamental and saturated through-fault error, around 8–10 pu of TAP.
- REF covers near-neutral winding faults where 87T sensitivity collapses as x².
- A grounding bank inside the 87T zone makes the LV side a zero-sequence source: remove I₀ on that winding too, and use the bank neutral CT for LV REF.
Knowledge check
- Q1A 230/34.5 kV MPT uses 1200:5 HV CTs and 3000:5 LV CTs. What is TAP1 on a 100 MVA base?
- Q2An OLTC regulates +10% / −15% around nominal, and the CTs may contribute up to 10% ratio error. What is a defensible SLP1?
- Q3Why does SLP2 need to be much steeper than SLP1?
- Q4During energisation, phase B shows only 11% 2nd harmonic while A and C show 24% and 17%. PCT2 is 15%. What prevents a trip?
- Q5What signature distinguishes overexcitation from inrush?
- Q6Where should the unrestrained 87U element be set?
- Q7Why does REF (87N) out-perform 87T for a ground fault 10% from the neutral of a grounded-wye winding?
Commission a secure, dependable 87T
Using the labs, produce a complete 87T / 87N / 24 setting set for the 230/34.5 kV, 100/133/167 MVA collector MPT from Step 1, with a +10% / −15% OLTC. Your target in the operate–restraint lab is 8 of 9 scenarios correct. The ninth, the near-neutral fault, is impossible for 87T by design and must be covered by REF.