Directional and Intermittent Earth-Fault Protection in Petersen-Coil-Grounded MV Networks

In a Petersen-coil-grounded MV network, successful compensation intentionally makes the fundamental earth-fault current small—the exact condition that defeats ordinary residual overcurrent protection. Selectivity must therefore come from the direction and character of the residual active, reactive or transient energy, with measurement errors, detuning and changing feeder capacitance treated as part of the protection problem.

This practical guide explains compensated-network physics, wattmetric and admittance methods, transient and intermittent-earth-fault algorithms, sensor requirements, settings, logic, commissioning and event diagnosis. It is vendor-neutral: parameter names, sign conventions and operating polygons must be taken from the selected relay manual.

Executive rules

Model total and feeder-by-feeder phase-to-earth capacitance in every switching state.
Define residual quantities explicitly: I_res = 3I₀ and U_res = 3U₀, or the relay’s alternative convention.
Measure small residual phase angles accurately; a few degrees of CT/VT error can dominate the active component.
Do not set feeder pickup from arc-suppression-coil current—the feeder sensor sees its own residual combination.
Use neutral displacement to detect system disturbance and a directional/multicriteria element to identify the faulted feeder.
Use a dedicated intermittent/transient algorithm for restriking faults that do not persist long enough for stable phasors.
Test under/overcompensation, high fault resistance, topology changes, switching events and forward/reverse faults dynamically.
Coordinate alarm, trip, fault-location and permitted first-fault duration with the network operating philosophy.

1. Why the Petersen coil changes protection

During a phase-to-earth fault, healthy phases/feeders supply capacitive current to the fault. A neutral reactor—arc-suppression or Petersen coil—supplies an opposing inductive current. With consistent per-phase definitions:

I_C ≈ 3ωC_0,totalV_phase, I_L = V_phase/(ωL), and at ideal tuning L ≈ 1/(3ω²C_0,total).

Real networks never reduce fault current to mathematical zero. Coil copper/core losses, network insulation losses, any parallel resistor, detuning, harmonics, asymmetry and the fault itself leave a residual active/reactive current. Compensation can help a transient arc self-extinguish near current zero, but it also makes fundamental-current magnitude a weak indicator of the faulted feeder.

2. Tuning, detuning and damping—state the convention

A project may define compensation k = I_L/I_C and detuning v = (I_C − I_L)/I_C. Under that explicit convention, v > 0 is undercompensated and v < 0 overcompensated. Other manufacturers reverse the sign or report percentage compensation. Never transfer a numeric setting without the equation.

Network damping: the active-loss component relative to capacitive current affects neutral voltage and wattmetric sensitivity.
Topology: cable/line switching changes C_0,total and each feeder’s contribution. The tuning controller must follow it within an approved time and range.
Parallel resistance: an auxiliary neutral resistor can intentionally increase active current for selectivity; include its duty and control state.
Frequency: both capacitive and inductive currents vary with frequency in opposite ways; protection and tuning must track the permitted range.
Asymmetry: transposed assumptions rarely describe an MV cable network perfectly; record healthy residual baseline by topology.

3. What each feeder relay measures

The source neutral/coil current is a system quantity. A feeder CBCT measures the vector sum of conductors passing through its window. On healthy feeders during an earth fault elsewhere, residual current is principally that feeder’s capacitive contribution toward the fault. On the faulted feeder, the measured residual is the vector balance of the return contributions and fault path, not simply “total charging current” or “coil current.” Build relay-setting phasors from the feeder-oriented zero-sequence model.

Quantity	Purpose	Typical vulnerability
U_res (3U₀)	Neutral displacement/start/polarisation	VT fuse/phase error, ferroresonance, small asymmetry baseline
I_res (3I₀)	Feeder residual magnitude and angle	CBCT ratio/phase error, screen routing, noise and dynamic range
P_res = Re{U_resI_res*}	Active/wattmetric direction	Very small in compensated networks; sign/scaling convention
Y_res = I_res/U_res	Admittance/conductance/susceptance discrimination	Unstable when U_res is below threshold; topology dependence
Transient charge/energy	Fault-inception/restrike direction	Switching/noise aliasing, sampling and time alignment

The power equation above is conceptual. Relays may use U₀/I₀, 3U₀/3I₀, rotated quantities, RMS windows or proprietary normalisation, so displayed watts and forward sign need not match another device.

4. Fundamental-frequency directional methods

Method	Operating principle	Engineering fit	Risk to test
Zero-sequence angle (“phi”)	Classifies I_res angle relative to U_res	Clear directional separation with known grounding state	Angle changes with detuning; CT/VT errors near boundary
Wattmetric/conductance	Uses active residual current/power direction	Resonant networks where active loss/fault component reverses on faulted feeder	Small operating quantity and parallel resistor/coil-loss allocation
Varmetric/susceptance	Uses reactive residual component	Some isolated/detuned applications	Direction may change across resonance
Admittance polygon	Evaluates G/B region in residual admittance plane	Wide topology/resistance range when modelled and commissioned	U_res threshold, healthy feeder capacitance and relay convention
Multicriteria	Combines magnitude, angle, harmonics, transient or persistence	Difficult networks and security improvement	Opaque vendor logic unless settings and test evidence are controlled

5. Wattmetric protection in practical terms

Near compensation, reactive current is largely cancelled and the residual active component becomes valuable for direction. The relay commonly rotates or projects residual current on the residual-voltage axis and applies a directional sector plus minimum U_res/I_res or active-current threshold.

Calculate the minimum forward active component at maximum fault resistance and every coil/resistor state.
Calculate the maximum reverse/healthy active component from feeder dielectric loss, sensor phase error, standing unbalance and noise.
Set the operating threshold between those populations with documented tolerances; if no gap exists, improve measurement or use another criterion.
Map relay “forward” to current flow and CT/VT polarity on the as-built one-line. Verify with phasors, not labels.
Supervise with neutral displacement and define reset/dropout/persistence so momentary switching does not trip.
If a parallel damping resistor is switched, use status/settings groups or a method proven secure in both states.

6. Why intermittent earth faults need a dedicated element

A deteriorating cable termination or insulation defect may arc for only a fraction of a cycle, extinguish and restrike after dielectric recovery. Each pulse produces high-frequency/transient residual current and neutral-voltage displacement, but a fundamental phasor element may reset between pulses or average the event below pickup. Waiting for stable 67N can therefore miss a damaging repetitive fault.

A dedicated intermittent-earth-fault (IEF/IGF) algorithm may detect correctly directed residual pulses, integrate energy/evidence or count qualifying events inside a rolling time window. For example, an application may assert only after N directional events in T seconds—but N, T, polarity, amplitude, dead time and reset are product-specific and must be established from network disturbances and test files.

Use U_res or another system-disturbance criterion to arm the event detector.
Require directional correlation between voltage and current transient where the relay supports it.
Set pulse count/time to detect insulation breakdown before escalation yet reject capacitor, reactor, transformer and cable switching.
Decide alarm-first versus trip based on asset criticality and permitted earth-fault operation.
Capture high-sample-rate oscillography and disturbance records; slow SCADA RMS trends are insufficient.
Do not confuse a conductor high-impedance fault algorithm intended for downed overhead conductors with an intermittent compensated-network earth-fault element.

7. Sensor and wiring design

CBCT: specify low-current magnitude and phase accuracy, usable frequency response for transient functions, aperture, maximum phase-fault immunity and relay burden/cable length.
Phase-CT residual: three CT ratio/phase mismatch and saturation often set a much higher sensitivity floor; use only when the study and dynamic tests support it.
VT: verify primary connection can reproduce zero sequence; open/broken-delta winding rating and damping; secondary earthing and fuse supervision.
Cable screens: route all screen/earth returns consistently. If a screen earth lead passes back through the CBCT, it can cancel the very fault current being measured.
Analogue wiring: segregate, shield/earth as specified, avoid uncontrolled terminal resistance and test with the final cable/relay.
Digital sensors/SV: verify scaling, channel mapping, quality, time, network redundancy, delay and failure behaviour end to end.

8. Settings workflow

List all feeder/source/bus-coupler states and coil/resistor control states.
Calculate C₀ for each feeder and total network; validate with field tuning/neutral measurements where available.
Calculate or simulate U_res, I_res phasors/transients for faults on every feeder, bus and source at low/high resistance.
Apply CT/VT magnitude/phase errors, frequency, noise, healthy unbalance, coil-range/tuning and study tolerances.
Choose the relay method whose forward and reverse populations have a demonstrable margin.
Set U_res arming/pickup above maximum healthy/switching residual yet below minimum fault displacement.
Set directional/active/admittance threshold and sector from the exact relay equation and as-built polarity.
Set delay/persistence and IEF pulse count/window; coordinate with permitted first-fault duration and fault-location process.
Define VT failure, CBCT circuit, coil-controller failure, low-quality SV and settings-group behaviour.
Run dynamic model/playback tests and record accepted security/dependability margins.

9. Illustrative network calculation

Illustrative only. A 20 kV, 50 Hz network study gives uncompensated total capacitive current I_C = 120 A in one topology. The coil is set to I_L = 114 A. Under the convention v = (I_C − I_L)/I_C, detuning is (120 − 114)/120 = 5% undercompensated, leaving about 6 A reactive mismatch before losses, harmonics and fault resistance.

If the calculated active residual contribution is 8 A, the simple orthogonal magnitude is √(6² + 8²) ≈ 10 A. This is not a universal feeder pickup: individual feeder relays see feeder-oriented currents, sensor errors may be comparable with the 8 A active component, and a cable switching operation can change both 120 A and detuning. The example shows why 51N magnitude alone is fragile and why directional active/admittance models plus topology testing are necessary.

10. Protection and control logic

Start: neutral displacement and/or validated transient criterion.
Select: forward wattmetric/admittance/IEF evidence on one feeder; restrain reverse/healthy feeders.
Operate: alarm, delayed trip or accelerated trip by fault type, repetition and operating policy.
Bus/source fault: define behaviour when multiple feeder elements indicate reverse or no feeder selects; provide bus/transformer protection and backup.
Double earth fault: phase/differential/overcurrent may need fast escalation; do not wait for first-fault location.
Coil control: alarm out-of-range, drive failure, excessive neutral displacement and unsafe detuning; map state to settings groups only with fail-safe validation.
Reclose: apply only when asset/network practice permits; repetitive cable IEF is not automatically a transient overhead-line fault.

11. Dynamic commissioning matrix

Test family	Minimum cases	Evidence
Steady forward/reverse	Low/high R_f, under/tuned/overcompensated, minimum U_res	Pickup, sector, time, margins and phasors
Topology	Shortest/longest network, feeder/coupler changes, parallel resistor states	Correct group/tuning and selectivity
Intermittent	Pulse amplitude, polarity, spacing, count, extinction/restrike patterns	Alarm/trip count and reset behaviour
Security	Cable/capacitor/transformer switching, CT saturation, VT fuse failure, noise	No false select/trip; proper alarm/block
Complete path	Sensor through relay logic, breaker, BF and SCADA	Timing, targets, records and isolation

Use the final relay firmware and setting file/checksum.
Replay EMTP/RTDS or representative COMTRADE waveforms with correct channel scaling and time alignment.
Test adjacent healthy feeder relays simultaneously when possible; selectivity is a system property.
Verify live-load/healthy residual phasors and intentional coil movement after energisation.
Store oscillography settings at a sample rate/pretrigger adequate for IEF diagnosis.

12. Event-diagnosis checklist

Which topology, coil tap/position, controller mode and auxiliary resistor state existed?
Did U_res precede I_res, and did phasor signs match the configured forward convention?
Which feeder showed directed active/transient evidence; what did adjacent feeders show?
Was the event fundamental, transient, harmonic or a pulse train?
Did VT fuse-failure, CT saturation, SV quality/time or channel mapping alarms assert?
Was pickup close to a measurement/tolerance boundary?
Did tuning move before/during the event and change direction?
Were multiple phases/feeders involved, indicating a double earth fault?
Did the breaker clear and BF reset within measured current/aux-contact criteria?
Do records support insulation testing/fault location before re-energisation?

13. Frequent mistakes

Mistake	Consequence	Correction
51N copied from LRG feeder	Blind or nonselective	Compensated-network directional method
Coil current used as feeder current	Wrong pickup basis	Feeder-by-feeder zero-sequence model
Detuning sign undocumented	Direction/setting reversal	Record exact equation and state
CT magnitude class only	Active component corrupted by phase error	Specify low-current phase accuracy and complete chain
Steady sine injection only	IEF/security unproven	Dynamic waveform/playback matrix
Healthy feeder not tested	System selectivity unknown	Simultaneous forward/reverse verification
Topology changes ignored	Coil out of range/settings stale	Change trigger, tuning and group logic

References and further reading

Engineering note: The method names above describe families of algorithms, not interchangeable settings. Use the selected relay’s published equations, accuracy limits, firmware-specific manual and validated test model.