Surge Arrester Leakage Current Testing: What the Resistive Component Actually Tells You

By | August 31, 2026

A surge arrester spends almost its entire life doing nothing. Between operations it sits there as an insulator with a very small current flowing through it. That current is the only window you have into its condition while it stays in service.

The problem is that most of what you measure is capacitive current, not the resistive component that carries the condition information. Knowing which part matters, and what corrupts it, is the whole job.

The current has two parts and only one of them is useful

Clamp a meter on the earth lead of a healthy metal-oxide arrester at service voltage and you will typically read somewhere between 0.2 mA and 3 mA peak. Almost all of it is capacitive. It comes from the permittivity of the MO blocks themselves, from stray capacitance, and from grading capacitors if the design uses them. The specific capacitance of a metal-oxide resistor element is typically about 60 to 150 pF·kV/cm² at rated voltage, which is where that milliamp-range figure comes from.

The resistive part is much smaller. It runs about 2 % to 20 % of the capacitive current, so roughly 10 µA to 600 µA peak at +20 °C.

That small component is the one that carries information. When MO blocks degrade, the voltage-current characteristic shifts and the resistive current rises at the same voltage and temperature. The capacitive current does not meaningfully change when blocks deteriorate. There is no evidence it does. So a measurement dominated by capacitive current is mostly telling you about the arrester’s geometry, not its health.

Define the resistive component properly: it is the current at the instant the voltage peaks, where dU/dt = 0. At that moment the capacitive contribution is zero. Everything left is resistive.

Why the total leakage current reading disappoints you

Plenty of surge counters have a mA meter built in. Crews read it, write it in the log, and move on. It feels like condition monitoring. It mostly isn’t.

Two things kill the sensitivity. First, the resistive part is a small fraction of the total. Second, the two components are 90° apart in phase, so they add vectorially, not arithmetically. Combine both effects and the resistive current has to change enormously before the total shifts noticeably.

The result is that even a large increase in resistive current shifts the total leakage current only slightly. How slightly depends on the arrester’s own capacitive and resistive values and on whether you read peak, mean or r.m.s. — peak moves most, r.m.s. and mean move least. IEC 60099-5 reaches the blunt conclusion: total leakage current is not sensitive even to critical changes in the resistive component, so it is generally unsuitable as an indicator of gradual degradation. A block set can be well into trouble before your mA meter looks unusual.

Total leakage current has its uses. It confirms the arrester is connected, it flags gross faults, it costs nothing to read. As a trend indicator for gradual ageing, it is not the tool.

The four things that will corrupt a good measurement

Before choosing a method, know what fights you in the field.

Harmonics in the system voltage. This is the big one. Voltage harmonics push capacitive harmonic currents through the arrester that are comparable in size to the harmonic currents the non-linear resistance produces. If your method reads harmonics, it cannot tell the two apart without help.

Temperature. Resistive current is strongly temperature dependent. Across the normal ambient range of −10 °C to +40 °C, the correction multiplier swings by well over a factor of two. A reading taken on a July afternoon and one taken in February are not comparable numbers.

Applied voltage. Resistive current also climbs steeply with voltage in this region. Service voltage drifts. Two measurements at 0.68 U<sub>r</sub> and 0.74 U<sub>r</sub> are not the same test.

Surface leakage. Rain, humidity and pollution put current on the outside of the housing. Moisture ingress puts it on internal surfaces. Both mix into what you measure at the earth terminal. Measure in dry conditions, or bypass the external surface current to earth deliberately.

There is a fifth issue that only bites on tall columns. Stray capacitance and nearby equipment make the voltage distribution along the arrester non-uniform. The blocks at the earthed end may see a voltage that differs from the arrester’s overall voltage in both magnitude and phase. Two consequences follow. The resistive current measured in the earth connection may differ from the average resistive current along the column. And any method that uses the voltage across the complete arrester as its phase reference inherits an error from that phase shift.

The measurement methods, grouped by how they work

IEC 60099-5 sorts the practical methods into three families. The distinctions matter because they determine what hardware you need at the site.

Family A — extract the resistive current directly.

  • A1 uses a voltage reference to read the current at the voltage peak. Standard laboratory practice. On site the reference signal is the obstacle.
  • A2 uses a voltage signal to compensate the capacitive component with a bridge. More sensitive than A1. Still needs the reference.
  • A3 builds a synthetic reference from the leakage current itself, so no voltage signal is needed. Trades that convenience for sensitivity to voltage harmonics.
  • A4 sums the leakage currents of all three phases so the capacitive components cancel. Elegant, but it only works when the three capacitances are genuinely symmetric. In practice that means well-controlled geometry such as GIS arresters.

Family B — read the harmonics instead.

  • B1 measures the third harmonic of the leakage current. In the upturn region of the V-I curve, the third is the largest harmonic component of the resistive current, typically 10 % to 40 % of it. No voltage reference needed, which is why this method is everywhere. The weakness is exactly the one described above: even a few percent of third harmonic in the system voltage can produce very large errors in the evaluated harmonic current, with the size depending on phase angle, arrester capacitance and the shape of the V-I characteristic.
  • B2 fixes that. A field probe at the base of the arrester picks up a signal proportional to the capacitive third harmonic, which is scaled and subtracted. What remains is the harmonic from the non-linear resistance. This is the method to specify for in-service work.
  • B3 uses the fundamental component instead, obtained by filtering and integration. Harmonic immunity is excellent, but you are back to needing a voltage signal, usually from a VT secondary.

Family C — measure power loss. Integrate the product of instantaneous voltage and current over a cycle. Multiplication and integration heavily suppress the influence of voltage harmonics. Typical loss values run 5 mW/kV to 300 mW/kV, referred to rated voltage, at U<sub>c</sub> and +20 °C. Needs a voltage signal.

Two patterns are worth pulling out of that list. Every method that avoids needing a voltage reference pays for it in harmonic sensitivity, except B2, which buys the immunity back with a field probe. And phase shift error hits every method that relies on the complete arrester voltage as its phase reference — which is why long columns need extra care regardless of the technique.

Normalising before you compare

A raw number in a logbook is close to worthless. The reading only becomes a trend when you convert it to reference conditions.

IEC 60099-5 gives a service voltage of 0.70 U<sub>r</sub> and an ambient of +20 °C as an example of standard operating conditions. It is an example, not a mandate — what matters is choosing one reference set and staying with it. Manufacturers can supply correction multipliers as curves against U/U<sub>r</sub> and against ambient temperature, covering roughly 0.40 to 0.90 U<sub>r</sub> and −10 °C to +40 °C. Resistive current, third harmonic current and power loss each get their own multiplier, though the voltage and temperature behaviour of all three is broadly similar.

So the field procedure has three parts, not one:

  1. Record the leakage current quantity.
  2. Record the actual service voltage and the ambient temperature at that moment.
  3. Convert to the reference condition before comparing with anything.

Skip step two and step three cannot happen. This is the most common failure in arrester monitoring programmes.

One caution on temperature: the correction assumes a certain over-temperature of the blocks relative to ambient. The real varistor temperature can sit higher than that assumption after energy absorption, or under strong solar radiation. The correction is good for routine trending, not for diagnosing an arrester you already suspect is hot.

Setting the limit

Manufacturers can also give limits for the normalised quantity, expressed as an absolute value, as a relative change over time, or both. Ask for them at the tender stage. The data typically requested from a manufacturer covers selection parameters — rated voltage, U<sub>c</sub>, nominal discharge current, protective levels, short-circuit rating, mechanical loads — and diagnostic reference data is not usually in that package unless you ask. It is far easier to obtain before the order than five years into service.

Where no manufacturer limit exists, a comparative approach still works. Compare identical arresters of the same type and vintage on the same busbar under the same conditions. The outlier is the one to investigate.

The same logic underpins the simplest on-line check available. Thermal imaging works because arresters normally sit close to ambient. Compare nearby units of similar type and vintage; a difference greater than about 10 K on one of them is a reason to look closer. It tells you nothing about slow degradation, but it catches trouble that has already started.

Practical requirements at the site

Almost none of this works on an arrester without an insulated earth terminal. That single design decision, made at specification time, determines whether the unit can ever be monitored properly. The terminal also needs enough withstand voltage to handle the inductive drop that appears between it and the earthed structure during a discharge.

Permanently installed monitors have to survive operational and short-circuit stress. Portable instruments have to be safe to connect with the arrester energised. Both points deserve more thought than they usually get.

And when a normalised reading crosses a limit, talk to the manufacturer before acting. These measurements are easy to misread, and the conversion from a harmonic current to a resistive current level in particular can introduce large errors on its own.

FAQ

Can I test a surge arrester with a megohmmeter?

Not usefully for condition assessment of the MO blocks. An insulation resistance test applies d.c. and tells you very little about the a.c. resistive current behaviour that indicates degradation. It can still reveal gross problems such as heavy moisture ingress or a failed unit.

Is a rising leakage current always bad?

Not until you have normalised it. Higher ambient temperature and higher service voltage both raise the reading on a perfectly healthy arrester. Convert to reference conditions first, then judge.

Why does my third harmonic reading jump around between visits?

Most likely the harmonic content of the system voltage changed. Method B1 cannot separate capacitive harmonic current caused by voltage distortion from resistive harmonic current generated by the arrester. If the network has significant distortion, use a compensated instrument.

Do I need to de-energise the arrester?

No. On-line measurement at service voltage is the normal approach and the most common in practice. Off-line measurement with a separate source is used where laboratory-grade accuracy is needed or where site conditions make on-line access impractical.

What about surge counters — are they worth reading?

Yes, but for a different purpose. A counter tells you about discharge activity, not arrester condition. It may also miss impulses arriving less than 50 ms apart, and some designs will not register the short currents typical of metal-oxide arresters at all.

Does the capacitive current ever indicate a problem?

Not for block degradation. It is set by geometry and material permittivity, and there is no evidence it shifts meaningfully as the V-I characteristic deteriorates. A large unexplained change in capacitive current points to something structural, such as a moisture path or a shorted section, rather than normal ageing.

Author: Zakaria El Intissar

Zakaria El Intissar is an automation and industrial computing engineer with 12+ years of experience in power system automation, electrical protection, and SCADA systems. He founded InsulationTesting.com to share practical, field-tested guides on insulation resistance testing, equipment, and industry standards. His writing is used by electricians and maintenance engineers worldwide. Based in Morocco.

Leave a Reply

Your email address will not be published. Required fields are marked *