Temporary Overvoltage and VLF Cable Testing

Specify the source in microfarads per frequency, not in kilovolts.

The track ends where the laboratory meets the power system. Two long-duration stresses remain, and both require everything that came before. Temporary overvoltage is the link between the dielectric test, the surge arrester curve and insulation coordination. VLF at 0.1 Hz is the practical answer to a problem you have known since the AC generation module: testing kilometres of cable demands reactive power proportional to 2π·f·C·U², and the only negotiable variable in that equation is frequency.

There is a deliberate symmetry in this closing. The impulse module dealt with the fastest phenomenon in the laboratory, the 1.2 µs front. This one deals with the slowest: temporary overvoltage lasting from a few cycles to hours, and a withstand test at 0.1 Hz in which a single cycle takes 10 seconds. Between those extremes the physics changes character. Under impulse, voltage distribution is dictated by capacitance; at power frequency, by permittivity; at very low frequency and in direct current, conductivity — strongly dependent on temperature and field — begins to govern. That is why the methods in this module have their own rules, their own tables, and above all their own limits of interpretation.

What you will be able to do

• Explain the physical origin of temporary overvoltage — earth fault with its earth fault factor, load rejection, the Ferranti effect, resonance and ferroresonance, loss of grounding, and distributed generation — and characterise it by amplitude in per unit, duration, frequency and the energy previously absorbed by the arrester.

• Justify VLF physically: demonstrate that test reactive power falls in direct proportion to frequency, a factor of 600 between 60 Hz and 0.1 Hz, and delimit honestly the limits of the method's representativeness.

• Conduct VLF cable testing per IEEE Std 400.2-2024 — the voltage table by class, ramps, 30 and 60 minute levels, monitoring — and combine it with partial discharge and tan delta at 0.1 Hz, applying the mean, tip-up and stability criteria.

• Locate faults by time domain reflectometry with calibrated propagation velocity, arc reflection and a surge generator, integrating pre-location and pinpointing.

• Document acceptance criteria and integrate the results into insulation coordination, closing the loop: stress, withstand, protection, test evidence.

TOV: physics, origins and characterisation

Temporary overvoltage is defined as an oscillatory power-frequency overvoltage, or close to power frequency, weakly damped or undamped, of duration much longer than switching overvoltage. It is characterised by four parameters: maximum RMS value, frequency including subharmonics and harmonics, duration, and the prior energy state of the arrester. The complete characterisation matters because the thermal withstand of the varistor depends on all four simultaneously.

The dominant origin of TOV in medium voltage networks is the phase-to-earth fault. When one phase goes to earth, the healthy phases rise relative to ground, and the ratio between the highest RMS phase-to-earth voltage of a healthy phase during the fault and the phase-to-earth voltage at the same point before the fault is the earth fault factor k. Its value depends on the zero-sequence to positive-sequence impedance ratios, on the fault impedance and on location, and therefore on the neutral grounding method. Solidly earthed with low zero-sequence impedance gives 1.1 to 1.3. Effectively earthed gives 1.4 or less, the usual criterion for arrester application. Resistance or reactance earthing gives 1.3 to 1.7. An isolated neutral in the ideal condition gives √3, about 1.732, because the healthy phases rise to phase-to-phase voltage against earth. Resonant earthing sits near 1.73 and can be exceeded by detuning and transients. Resonance or abnormal neutral displacement exceeds 1.73 and has to be determined by electromagnetic transient simulation.

Coordinating the arrester TOV curve

The arrester module introduced continuous operating voltage and rated voltage. TOV adds the third quantity: the RMS voltage withstandable for a defined time, dependent on initial temperature and on previously absorbed energy. Suppliers publish it normalised, and the axis matters — some catalogues normalise by rated voltage and some by continuous operating voltage. A representative published curve for a distribution arrester family, normalised by rated voltage and without prior energy, gives 1.300 at 0.1 s, 1.225 at 1 s, 1.150 at 10 s, 1.120 at 100 s, 1.090 at 1 000 s and 1.062 at 10 000 s. With prior energy duty the same points fall to 1.215, 1.155, 1.095, 1.055, 1.015 and 0.975. The familiar 1.15 at 10 s is a product value without prior energy, not a universal limit: after a discharge, the 10 s capability of that family drops to about 1.095 times rated voltage.

The physics of 0.1 Hz, and the trap below it

For a capacitive load under sinusoidal voltage, the capacitive current is 2π·f·C·U and the reactive power is 2π·f·C·U². The active power dissipated is only that reactive power multiplied by tan delta; the bulk of the demand is reactive and scales linearly with frequency. Between 60 Hz and 0.1 Hz the ratio is exactly 600: a VLF source delivers one six-hundredth of the current and reactive power a power-frequency source would need for the same cable at the same voltage.

IEEE 400.2 recognises two principal waveforms. Sinusoidal, with distortion below 5% and RMS equal to peak divided by √2, is preferred for diagnosis, because tan delta and phase-resolved PD analysis have conventional definitions. Cosine-rectangular holds nearly constant levels with fast polarity reversals whose rate of change resembles half a cycle at 50 or 60 Hz; there the RMS value approximately equals the peak, which is why the cosine-rectangular voltage table sits about √2 above the sinusoidal one in RMS terms. The two forms are not interchangeable in interpretation: do not compare sinusoidal tan delta with a cosine-rectangular result, and do not compare inception voltages between waveforms. The standard itself stresses that VLF results do not automatically correlate with 50 or 60 Hz.

Levels, embedded diagnostics and the four-hour campaign

The current edition keeps the three-level architecture — installation, acceptance and maintenance — with maintenance at about 75% of acceptance. Sinusoidal phase-to-earth values by class are: 5 kV class at 9, 10 and 7 kV; 8 kV class at 11, 13 and 10 kV; 15 kV class at 19, 21 and 16 kV; 25 kV class at 29, 32 and 24 kV; and 35 kV class at 39, 44 and 33 kV. An important methodological note: the real table is not a uniform multiple of U0. At 15 kV class the sinusoidal acceptance value of 21 kV is 2.42·U0; at 25 kV class, 32 kV is 2.22·U0; at 35 kV class, 44 kV is 2.18·U0. The kilovolt value in the table prevails over any rounded multiple.

The structural limitation must appear in every report: tan delta is a capacitance-weighted average, so a defective joint in 8 km of healthy cable is diluted and a low mean does not exclude a localised defect. That is why the method walks paired with PD, which locates, and with phase-to-phase and normalised historical comparison. Under VLF the apparent charge is still calibrated in picocoulombs, but the physics at 0.1 Hz is not identical: inception, extinction, repetition rate, phase of occurrence and interface charge accumulation can all differ, and in 30 minutes there are 180 cycles against 108 000 at 60 Hz. The rules: never convert picocoulombs from 0.1 Hz to 60 Hz; never compare inception voltages between waveforms; keep source, sensor and bandwidth constant when trending; and treat location and evolution with voltage as more informative than any isolated limit.

Alternatives, fault location, and closing the coordination loop

A resonant hipot at 20 to 300 Hz uses a tuned reactor to compensate the capacitive reactance of the cable; at resonance the source feeds only the losses, and the quality factor multiplies the voltage. It is the reference method for high and extra high voltage cable: IEC 60840 and IEC 62067 provide for the after-installation test typically at 1.7·U0 for 1 h at 20 to 300 Hz, with lower frequencies or the alternative of U0 for 24 h subject to agreement and class. Its decisive advantage over VLF is representativeness, because PD measured at tens or hundreds of hertz compares directly with service.

When the withstand does its job and the cable fails in test, or in service, the problem changes from assess to find. A time domain reflectometer injects a low voltage pulse and reads the echoes from impedance discontinuities, with a reflection coefficient of +1 at an open circuit, −1 at a short, and zero at a matched termination. Distance follows from d = v_p·Δt/2, where the factor of two is the round trip. Propagation velocity in power XLPE is about 150 to 170 m/µs, a velocity factor of 0.50 to 0.57, dependent on effective permittivity, geometry and the semiconductive layers. In the 8 km example the echo from the remote terminal returns in 94 to 107 µs, and an error of only 5% in the assumed velocity displaces the fault by about 400 m: calibrating the velocity against a known length or a mapped joint is a mandatory step, not a refinement. The limit of plain TDR is that high-resistance faults, which are most faults in XLPE, are invisible to a low voltage pulse. The solution is arc reflection: a surge generator ionises the fault into a temporary low-impedance arc, and the TDR, synchronised through a filter, compares the trace during the arc against a reference trace, with the difference marking the fault. Use the minimum energy that stabilises the arc, because excessive impacts carbonise the channel, damage recoverable joints and create new defects. Final pinpointing is acoustic or electromagnetic over the pre-located point.

Syllabus

• TOV defined: oscillatory power-frequency overvoltage, weakly damped, characterised by four parameters

• The earth fault factor k by neutral grounding method, and TOV duration set by protection

• Loss of grounding: how an effectively earthed network temporarily becomes isolated

• Load rejection, the Ferranti effect, linear resonance and ferroresonance

• The arrester TOV curve with and without prior energy, and the five-step selection procedure

• The Ur trade-off: thermal robustness against protective level

• TOV in the transformer: the induced voltage test and current terminology

• Why frequency must rise and duration must compensate in cycles

• The physics of 0.1 Hz: reactive power, the worked example of 8 km, and the stored energy safety warning

• Sinusoidal against cosine-rectangular waveform and why they are not interchangeable

• Frequencies below 0.1 Hz and the cycle-count trade-off

• IEEE 400.2-2024 levels by class for installation, acceptance and maintenance

• Durations and the logic of the withstand test

• Embedded diagnostics: tan delta at 0.1 Hz with mean, tip-up and stability criteria by insulation family

• Partial discharge under VLF and what must not be converted

• TDR, arc reflection and pinpointing

• Alternatives: resonant hipot, damped AC, and why DC died for extruded cable

• Insulation coordination: the four conditions the arrester must satisfy simultaneously

Laboratory work

Directed study: you have a 4 hour outage window and 8 km of cable, so design the campaign. The exercise demands a complete test plan covering VLF, tan delta, partial discharge and time domain reflectometry with criteria and evidence. Premises: XLPE of 15 kV class, U0 = 8.66 kV, C of 1.6 to 3.2 µF per phase, aged cable, maintenance objective, sinusoidal level of 16 kV RMS, 30 minutes per phase at 0.1 Hz. A representative schedule: 0:00 to 0:20 for safety and configuration — lockout and tagging, absence of voltage, grounding, disconnection of arresters, voltage transformers and loads, and definition of the test zone, with no branch connected and the shielding scheme verified. 0:20 to 0:40 for preliminaries — conductor and shield continuity, sheath resistance, auxiliary insulation resistance, and real capacitance measurement per phase, with any capacitance divergence between phases above 5% investigated before energising. 0:40 to 1:00 for reference TDR on all three phases, with the remote terminal echo, velocity calibration and a joint map, archiving raw traces with scale. 1:00 to 1:45 for stepped tan delta and PD at 0.5, 1.0 and 1.5·U0 with at least six readings per level, extracting mean, tip-up and stability plus inception voltage, pattern and location — this is the decision gate, and any parameter in the action band or intense localised PD means the withstand is not applied. 1:45 to 3:15 for the maintenance VLF: 30 minutes on each of three phases at 16 kV RMS with current and tan delta monitoring, discharging and grounding between phases, with all events time-stamped. 3:15 to 3:35 for post-test work: repeat tan delta at U0 and TDR on any anomalous phase, compare before and after, confirm zero residual charge and recheck sheath resistance. 3:35 to 4:00 to close out: classification per phase, preliminary result, restoration of arresters and terminations, release, and the baseline archived for trending.

Key takeaway

Measure the capacitance before you mobilise. A VLF source is specified in microfarads per frequency, not in kilovolts, and 36 cycles are not 180. Use tan delta as the gate before the withstand, and leave the field with a baseline rather than a pass or fail.

Module 23 of the Atlas Energy Academy · Diagnostic Methods · about 4 hours · Two short video series plus a directed case study · Prerequisites: Modules 18 and 20