On an ungrounded or high-resistance-grounded medium-voltage network, a sudden jump in zero-sequence voltage tells you something changed. It doesn't tell you what. A single phase-to-ground fault and a ferroresonant oscillation can look surprisingly alike at the bus-especially when the ferroresonance is fundamental-frequency type. A locating device that acts on zero-sequence voltage magnitude alone will mix them up. And that mistake isn't harmless. A ferroresonance event gets treated as a persistent ground fault, or a real ground fault gets waved off as resonance that will settle down. Either way, you're looking at unnecessary switching, or prolonged overvoltage stress on voltage transformers and cables.
Timing is a good place to start, and it's usually the most practical thing to put into firmware. A single phase-to-ground fault is a temporary overvoltage condition. It lasts until the fault is cleared or the section is isolated. Ferroresonance, once excited, also tends to persist, but its onset behaves differently. It can be triggered by a switching operation or by clearing a ground fault, and the move into the resonant state isn't instantaneous. What matters for a locating device is the opening window. In the first 0.1 s after the disturbance, a single phase-to-ground fault shows an essentially power-frequency component at 50 or 60 Hz, with little else worth talking about. A ferroresonant event, on the other hand, produces a cluster of adjacent frequency components-subharmonics, integer harmonics, or a mix-inside that same early window. A device that captures and analyzes the first few cycles can use that frequency spread to separate the two before either has had time to evolve.
Waveform shape matters when the ferroresonance is fundamental-frequency type. In that mode the voltage waveform can look deceptively close to a ground fault in RMS and phase-angle terms. The difference is distortion. A clean single phase-to-ground fault leaves the zero-sequence voltage waveform largely sinusoidal, apart from transient ringing at inception. Fundamental ferroresonance carries a distorted zero-sequence waveform because the saturable magnetizing inductance of the voltage transformer is already working on a nonlinear part of its characteristic. A useful auxiliary check is to fit the measured zero-sequence voltage to a 50/60 Hz sinusoid and compare the fitted amplitude with the actual peak. That ratio-or a waveform-distortion index derived from it-rises when nonlinear resonance is present and stays near unity for a clean ground fault. You don't need a full harmonic spectrum for this. You need enough sampling resolution to see the shape, not just the magnitude.


Zero-sequence current is where locating devices earn their keep. Voltage features alone are ambiguous in the fundamental ferroresonance case, but the current path tells a different story. During a single phase-to-ground fault, the zero-sequence current in the faulted section is dominated by the system's capacitive ground current and flows with a definable relationship to the zero-sequence voltage. During ferroresonance, the zero-sequence current is governed by the nonlinear excitation branch of the voltage transformer and doesn't follow the same phase or amplitude pattern. A device that measures both residual voltage and residual current at each monitored point can apply a directional or phase-relationship check as an auxiliary criterion, rejecting ferroresonance even when the voltage signature is ambiguous. Transient zero-sequence current polarity has also been used effectively for ground-fault section identification, and the same measurement chain supports ferroresonance discrimination without adding hardware.
For an independent-site device, the practical implication is simple enough: don't lean on a single threshold. The device should record triggered waveforms at a sampling rate high enough to preserve the first 100 ms of the event, compute the frequency spread in that window, check the zero-sequence voltage waveform for sinusoidal fit error, and-where a zero-sequence current channel is available-compare its phase or polarity against the voltage. A locating device that logs only RMS values or one-cycle aggregates will struggle to separate fundamental ferroresonance from a ground fault. A device that captures raw samples around the trigger and runs these checks in combination gives the operator a defensible classification rather than a guess. The point isn't to replace protection logic with a diagnostic. It's to make the diagnostic good enough that protection and operations decisions aren't built on a misread waveform.