Metal-Oxide Surge Arrester Testing

The only device in the substation designed to conduct when everything goes wrong.

Every other piece of equipment in a substation is tested to withstand voltage without conducting. The metal-oxide surge arrester is tested to conduct tens of kiloamperes and then, with the discharge over, to return immediately to a quiet, thermally stable state under system voltage. The zinc oxide varistor that makes this possible is one of the most remarkable nonlinear materials in electrical engineering: seven decades of current accommodated with roughly 50% variation in voltage.

That permanent electrical continuity with the system is both the elegance and the Achilles heel of the design. The arrester is always energised, always conducting a small leakage current, always warming. Every test in this family circles, directly or indirectly, around the thermal equilibrium of that material. The central question running through the whole test programme is a single one: does the arrester absorb the surge energy and return, stable, to steady state? This module is short, dense, and built entirely on impulse current generation and the general test taxonomy established earlier in the track.

What you will be able to do

• Explain ZnO varistor physics — grain microstructure, intergranular barriers, the nonlinearity coefficient α, and the regions of the V-I curve — and relate it to behaviour in service and in test.

• Define and distinguish Ur, Uc (MCOV), Uref with its reference current Iref, Ures and the TOV curve, and extract those quantities from real data sheets.

• Run and interpret the routine tests of IEC 60099-4:2014 — reference voltage, residual voltage, internal partial discharge at 10 pC or less, tightness, and current sharing between parallel columns.

• Describe the complete type test architecture: residual voltage curve (steep, 8/20 µs and 30/60 µs), long-duration impulses, high current 4/10 µs, operating duty, thermal stability, 1 000 h accelerated ageing, short circuit and pressure relief, artificial pollution, RIV, bending moment and disconnector.

• Convert between the historic line discharge classes 1 to 5 and the current Qrs, Qth and Wth classification (SL/SM/SH and DL/DM/DH) when reading old and new catalogues side by side.

• Apply the diagnostic methods of IEC 60099-5:2018 — total leakage current versus resistive component, and the compensated third-harmonic method — and interpret TOV curves with and without prior energy.

Seven decades of current, fifty percent of voltage

A varistor block is not a homogeneous piece of zinc oxide. It is a polycrystalline ceramic produced by milling, doping, pressing and sintering, in which semiconducting ZnO grains, typically 5 to 15 µm across, are separated by an intergranular layer rich in bismuth oxide and by spinel phases that control grain growth. Each grain boundary behaves as a double Schottky barrier of roughly 3.2 to 3.5 V, and the total block voltage is the sum of the drops across thousands of boundaries in series. That architecture produces the nonlinearity coefficient α of 25 to 60 in the protection region: in the model I = k·V^α, a current ratio of 10⁷ from 1 mA to 10 kA is accommodated, for α = 40, by a voltage ratio of only about 1.50. High-gradient materials shrink the grain to around 5 µm, increasing the number of barriers per millimetre and with it the operating field.

Design numbers follow directly from that microstructure. The residual gradient at 10 kA, 8/20 µs, typically sits between 300 and 350 V/mm in large blocks of about 100 mm diameter, 340 to 400 V/mm in intermediate blocks of about 60 mm, and 400 to 500 V/mm in distribution blocks of about 32 mm. A 60 mm × 36 mm block sustains roughly 13 kV during a 10 kA discharge, at a current density near 379 A/cm². The effective relative permittivity of the ceramic is on the order of 600 to 1 100, giving a typical block 0.5 to 2 nF of capacitance and a complete column tens to a few hundred picofarads — a value that interacts with stray capacitance to earth and forces the use of grading rings on tall extra high voltage arresters.

Five quantities that specifiers routinely confuse

Rated voltage Ur is the highest power-frequency RMS value the arrester must withstand temporarily under the conditions of the operating duty test, typically applied for 10 s after energy injection. It is not the system nominal voltage, it is not the maximum system voltage, and it is not continuously applicable. Continuous operating voltage Uc, called MCOV in the IEEE tradition, is the highest RMS value applicable permanently; the ratio Uc/Ur typically falls between 0.75 and 0.84, with 1/1.25 = 0.80 the most common.

Two patterns are worth carrying in your head. Switching residual voltage typically sits at 75 to 90% of the residual at 10 kA, 8/20 µs, so it is the lowest point of the protective curve. And doubling the impulse current raises Ures by only 5 to 15% depending on the point on the curve, which is the direct consequence of the high α. Small distribution arresters tend toward higher Ures/Ur ratios, sometimes above 3, while heavy multi-column extra high voltage designs approach 2. The ratios must always be read from the data sheet, never estimated from a single multiplier.

Operating duty: the proof of thermal recovery

The operating duty test is the synthesis test of the arrester. Initial Uref and Ures are measured; the sample is preconditioned where prescribed with high current impulses; it is heated to a starting temperature representative of service, around 60 °C; it receives the specified energy injection — for example three long-duration impulses of 2 to 4 ms, each carrying about one third of Wth, at intervals under 60 s; within less than 100 ms after the last impulse, Ur is applied for 10 s; the voltage is reduced to Uc for 30 minutes; and temperature, resistive current and power are monitored throughout.

Accelerated block ageing complements the duty test on a longer timescale. Block samples are held at 115 ± 4 °C, energised at 1.05·Uc or the corrected ageing voltage, for 1 000 hours, with continuous monitoring of dissipated power. The classic criterion is that final power must not exceed initial power, demonstrating that the material loss curve is stable or falling over life. If power rises, the test determines correction factors that raise the voltages used in the other thermal tests, penalising the unstable design. This must not be confused with the 1 000 h weather ageing of a polymeric housing under salt fog, which has its own criteria of no tracking or erosion, Uref reduction within about 5%, and PD at or below 10 pC at the end.

Failing without exploding, and the electrocardiogram of the varistor

When an arrester fails — and terminal failure is always a short circuit of the column — system fault current flows through the housing until protection operates. Without a tested pressure relief design, a porcelain-housed arrester is a ceramic grenade: the internal arc vaporises material, pressure rises in milliseconds, and fragmentation throws debris. The short-circuit test of IEC 60099-4 deliberately provokes the failure with a shorted column and applies the declared rated short-circuit current — standardised values of 80, 63, 50, 40, 31.5, 20, 16 or 10 kA RMS, with an asymmetrical first peak of about 2.5 times the RMS value and a duration on the order of 200 ms — plus a low current test at 600 ± 200 A for about 1 s. The criteria are no dangerous fragment above 60 g outside the prescribed area, controlled opening of the pressure relief devices without violent rupture, and self-extinguishing of flames within about 2 minutes. This is a personnel safety test, not a performance test, and no economy justifies omitting it.

Annex D of IEC 60099-5:2018 catalogues the methods: direct measurement with a voltage reference or capacitive compensation, harmonic analysis, and direct power measurement. The most used in the field is third-harmonic analysis of the leakage current with compensation for the third harmonic already present in system voltage — synchronised sampling, FFT, vector subtraction of the component created by network distortion, correction for temperature and voltage, and comparison against the baseline. Without that compensation, a simple change in supply voltage quality would be diagnosed as ageing.

IEC 60099-5 fixes no universal condemnation limit in milliamperes; the diagnosis is by trend. Established engineering heuristics: a corrected increase of 30 to 50% over baseline calls for investigation, approaching twice the baseline calls for priority assessment, and rapid growth between campaigns or abnormal localised temperature calls for immediate action. Combined with a surge counter, this makes the arrester the most monitorable device in the substation.

Syllabus

• Microstructure of the ZnO block: grains, intergranular barriers, α, residual gradient in V/mm

• The V-I curve and its four regions: leakage, knee, protection, upturn

• Thermal runaway: exponential generation against near-linear dissipation, and the two intersection points

• The five defining quantities: Ur, Uc/MCOV, Uref, Ures and TOV

• Why an applied voltage test does not apply to the active column, and what is tested dielectrically

• Routine tests: reference voltage as fingerprint, residual voltage by block, internal PD, tightness, current sharing

• Type tests: the full residual voltage curve and the definition of protective levels

• Long duration impulses, charge transfer, and the shift from line discharge class to Qrs/Qth/Wth

• High current impulse 4/10 µs as structural robustness, not energy

• Operating duty and the thermal stability acceptance criterion

• Accelerated block ageing at 115 °C and 1.05·Uc for 1 000 h

• Short circuit and pressure relief: failing without exploding

• Artificial pollution, RIV, bending moment (SLL and SSL), and disconnector type tests

• TOV curve with and without prior energy

• Field diagnostics: separating the resistive leakage component and interpreting trend

Laboratory work

Two routines using new distribution varistor blocks of about 32 mm and blocks artificially aged in an oven at 115 °C under 1.05·Uc. Routine 1 — reference voltage and leakage current: energise each block with rising voltage to the reference current (use Iref = 1 mA peak) and record Uref with the temperature noted; then at 0.8·Uref acquire u(t) and i(t) simultaneously for at least 10 cycles, run the FFT, and separate the capacitive fundamental in quadrature, the resistive fundamental in phase, and the third harmonic. Compare new and aged blocks: the aged one shows Uref a few percent lower and visibly higher resistive leakage. Repeat with the block heated to 60 °C and quantify the effect of temperature on the resistive component. Routine 2 — residual voltage with an 8/20 µs impulse: apply 0.5·In, In and 2·In, recording i(t) with a coaxial shunt and u(t) with a compensated divider. Plot Ures against I on log scale, compute the effective α between point pairs and confirm it lands in the 25 to 60 band, check repeatability between two consecutive impulses at In (Ures variation of 2% or less), and compare the Uref signature before and after — a change beyond about 5% indicates block damage. Safety note: keep the current loop area minimum, keep the operator out of the plane of the loop, and follow the full capacitor bank grounding procedure.

Key takeaway

Reference voltage is not a nameplate number, it is a vital sign. Two arresters with the same catalogue rating but 8% difference in measured Uref are telling you something — measure the resistive leakage of both before deciding where either one goes, and never put mismatched units in parallel columns of the same pole, where the difference in V-I curve concentrates discharge current in the lower-voltage unit.

Module 16 of the Atlas Energy Academy · Equipment Families · about 4 hours · Video series (8 × 10–15 min) plus bench practice · Prerequisites: Module 11 — Test Taxonomy (Module 8 recommended)