Selecting the faulted feeder in a Petersen-coil compensated network — the wattmetric and transient methods

The neutral voltage displacement protection guide (ANSI 59N) explains how a broken-delta voltage-transformer connection reliably detects that an earth fault has occurred somewhere on an ungrounded or resonant-earthed network, but explicitly notes that this detects the fault's existence, not its location. On a network earthed through a Petersen coil, locating which feeder is actually faulted is a distinct problem from the coil's own compensation function, and it is harder than it looks: the coil is deliberately tuned so that its inductive current cancels out the capacitive earth-fault current of the network, which is precisely what makes an ordinary residual-current or reactive-power directional relay (of the kind described in the 67 directional overcurrent guide) unreliable for feeder selection on this specific network type.

Why ordinary reactive-power direction fails once the coil is tuned

On a solidly or low-resistance earthed network, the 50N/51N earth-fault overcurrent guide and the ordinary 67 directional principle both work because the residual current at the fault location is large and its direction relative to the residual voltage clearly distinguishes the faulted feeder from healthy ones. On a Petersen-coil compensated network, the coil is tuned so that its own reactive (inductive) contribution to the fault current approximately cancels the network's capacitive contribution at the fault point. The near-complete cancellation that makes the coil effective at limiting the fault current and letting the network ride through a transient earth fault is exactly what removes the clean reactive-current signature a conventional directional relay would otherwise use to pick out the faulted feeder — a healthy feeder's own capacitive contribution and the faulted feeder's residual current can end up looking too similar in magnitude and phase for a simple reactive-power comparison to distinguish reliably.

The steady-state wattmetric method

Because the coil is tuned to compensate the network's reactive (capacitive) current, real protection schemes commonly exploit the resistive, in-phase component of the residual current instead — a function generally referred to as wattmetric or active-power earth-fault protection, and often labelled 67Ns or "sensitive earth fault" in relay documentation (the exact designation varies by manufacturer). A small resistive component is intentionally present in the fault-current path, either from a parallel damping resistor across the coil (sized to provide a defined, measurable active current) or from the network's own inherent conductance losses, and this resistive component flows in a characteristic direction at the fault location that differs from the direction seen at a healthy feeder:

On the faulted feeder, the measured active power (the in-phase product of the residual voltage 3V0 and the residual current 3I0) flows from the busbar towards the fault, in one defined direction.
On a healthy feeder, the active power associated with that feeder's own small conductance losses flows in the opposite direction relative to the busbar.

A wattmetric earth-fault relay compares this direction and magnitude of active power on each feeder and selects the one whose active-power direction is consistent with a fault, rather than relying on the reactive-power comparison that a conventional 67 relay uses on a solidly earthed network.

The transient (first-cycle) method

An earth fault does not begin as a clean steady-state condition: at the instant the fault occurs, the network's phase-to-earth capacitances discharge and recharge through a brief, high-frequency transient before the power-frequency, compensated steady state described above is reached. Because the Petersen coil is tuned to the network's power frequency, it does not meaningfully influence this initial high-frequency transient, which is dominated instead by the network's own capacitance and the fault's initiation instant. A transient (or "first-cycle") earth-fault detection method captures this brief event, in the first few milliseconds after fault inception, and compares its magnitude and polarity feeder by feeder: the faulted feeder typically shows a markedly larger transient discharge current than a healthy feeder, with a polarity that distinguishes it from the healthy feeders' own smaller capacitive-discharge contributions. Because this method relies on a fast, one-shot transient rather than a sustained steady-state signal, it depends on adequately fast sampling and triggering in the protection relay, and it is generally used either as the primary detection method or as a corroborating check alongside the steady-state wattmetric method, rather than as a completely independent, standalone alternative.

Why the two methods complement rather than replace each other

The wattmetric method depends on a real, measurable resistive component being present in the fault-current path, which in turn depends on the coil's parallel damping resistor (or the network's own conductance) being large enough to produce a reliably measurable active-power signal — a network with very low conductance losses and no parallel damping resistor can make the wattmetric signal too small to detect confidently. The transient method does not depend on this resistive component at all, but instead depends on the fault initiating with enough instantaneous discharge and on the relay sampling fast enough to capture it — an intermittent, self-clearing earth fault (common on overhead or aged cable networks) may produce a usable transient event on every restrike without ever settling into a steady-state fault the wattmetric method could evaluate. In practice this is why many protection relays on resonant-earthed networks implement both principles side by side, rather than relying on a single method for every fault scenario.

Practical relevance

When commissioning or reviewing earth-fault protection on a network earthed through a Petersen coil, it matters to verify not just that the 59N displacement-voltage protection correctly flags that a fault exists, but also that a feeder-selective method — wattmetric, transient, or both — is actually commissioned and correctly set on every outgoing feeder, and that the coil's parallel damping resistor (if the wattmetric method is used) is sized to produce a large enough active-current signal for the relays actually installed. A network with a functioning 59N alarm but no working feeder-selective protection tells the operator that a fault exists somewhere, but leaves locating it to manual switching and fault-passage indicators.

Common mistakes

Assuming an ordinary 67 directional relay, tuned for a solidly earthed network, will correctly select the faulted feeder on a compensated network — the coil's own compensation removes the clean reactive-current signature that conventional directional relays depend on.
Sizing the coil's parallel damping resistor without considering the wattmetric relay's sensitivity — too small a resistive contribution can leave the active-power signal too weak for reliable feeder selection.
Relying on the wattmetric method alone on a network prone to intermittent, self-clearing (restriking) earth faults — a fault that never settles into a steady state may only be catchable by the transient method.
Treating a correctly functioning 59N alarm as proof that feeder selection also works — displacement-voltage protection confirms only that a fault exists somewhere on the network, not which feeder is faulted.