Shunt Reactor Switching

This page complements the capacitive-switching notes by covering the opposite type of duty: switching a mainly inductive load, where the current is usually small but the recovery voltage and reignition risk can be severe.

A shunt reactor is an inductive load used to absorb reactive power and help hold voltage down, particularly on long lines or cables under light load. Reactors are switched often — sometimes daily — to follow changes in system loading, so the breaker also needs good mechanical endurance.

The reactor current is usually relatively small compared with fault current, commonly of the order of a few hundred amperes, so the challenge is normally not thermal interruption capability but dielectric recovery and overvoltage control. The difficulty is that the breaker may force the small inductive current to zero before its natural zero — current chopping — trapping energy in the reactor and producing an overvoltage across the open contacts. Two effects dominate: current chopping and reignition.

The duty is therefore voltage-dominated rather than current-dominated: the breaker can usually interrupt the current, but it must then withstand the recovery voltage without reignition.

A breaker does not always wait for the natural current zero. With a small inductive current, arc instability forces the current prematurely to zero — this is called current chopping, and it can be treated as effectively instantaneous. The magnetic energy left in the reactor must then transfer into the load-side capacitance, producing an oscillatory overvoltage. The chopped current is set by the breaker’s chopping number and the capacitance in parallel with it; it is largely a property of the breaker and its interrupting medium. SF₆ self-blast and rotating-arc designs chop very little; air-blast and minimum-oil designs chop more.

For a single-interrupter breaker the chopped current follows from the chopping number \(\lambda\) and the capacitance in parallel with the interrupter; for \(N\) interrupting units per pole that capacitance scales with \(N\):

The chopping number characterises how strongly a breaker technology tends to chop small inductive currents. It is a property of the interrupter and applies to every type except vacuum, where the degree of chopping is set mainly by the contact material. Typical ranges span nearly two orders of magnitude:

After chopping, the reactor current stops abruptly. The reactor inductance and the load-side capacitance then exchange energy, typically at about 1 to 5 kHz: the first peak of that oscillation is the suppression peak overvoltage, and the following opposite-polarity peak contributes to the recovery voltage across the open breaker contacts. For a directly grounded reactor the recovery voltage peak is no greater than the suppression peak.

The size of the suppression peak follows from the chopped current and the reactor loop. In per unit of \(V_{o}\) (the crest voltage across the reactor at interruption) it is:

The right-hand form (valid when \(C_{s}\gg C_{L}\) and \(C_{p}\) is negligible) makes the key point: the chopping overvoltage depends only on the chopping number and the reactor rating — a smaller, lower-rated reactor switched by a high-chopping breaker gives the highest overvoltage. Because the load-side oscillation is only lightly damped, the opposite-polarity recovery-voltage peak is of nearly the same magnitude.

After the current is chopped, the voltage across the open contacts is the difference between the source-side voltage (near its crest) and the oscillating load-side voltage. If the contacts have not yet separated far enough to withstand that voltage, the gap breaks down again — a reignition.

Whether a reignition happens depends on the contact-parting instant relative to current zero and on how fast the gap regains strength (its rate of rise of dielectric strength). Reignitions create steep, fast transients — fronts from under a microsecond to a few microseconds. Because the front is so steep, the voltage does not distribute evenly along the reactor winding, so the entrance turns can see higher stress than the power-frequency voltage would suggest, and winding resonances can be excited.

A reignition drives the load-side voltage past the source voltage and overshoots. With damping, the reignition overvoltage peak to ground \(k_{p}\) and its peak-to-peak excursion \(k_{s}\) (both in per unit of \(V_{o}\)) are:

These are the directly grounded forms. An ungrounded reactor, or one grounded through a neutral reactor, raises the bracketed term (to \(2+k_{a}\) and \(1+2K+k_{a}\) respectively), which is why the grounding arrangement governs the severity (Section 5).

The figure marks the switching voltages directly; dividing each by the reference voltage \(V_o\) gives the per-unit factor used above:

A reignition is not a single-frequency event. After a clean interruption the load side rings slowly as trapped energy swaps between \(L\) and \(C_L\):

When a reignition does occur, the transient unfolds as a sequence of much faster oscillations, as different parts of the circuit discharge and re-share their charge at different frequencies. An EMT study has to capture the whole span — from the low-frequency reactor oscillation to the very fast breaker and capacitance transients, a few kilohertz to several megahertz — so the four modes below set the bandwidth and time-step the model needs.

This sequence is what links reignition to escalation (next section): if the breaker interrupts one of these fast currents, the load side is left charged and the next reignition can be larger. A further effect — beating of the recovery-voltage oscillation from phase-to-phase coupling, seen with longer breaker-to-reactor connections and three-phase (single-tank) reactors — does not change the first recovery-voltage peak, so it does not alter the reignition risk and is not a design concern here.

If a reignition is followed by a fresh interruption during the high-frequency oscillations, the load-side oscillation restarts. Each interruption can leave the reactor side charged to a more severe voltage, so the next reignition starts from a worse initial condition and can be larger than the last. Repeated, this is voltage escalation.

It is most associated with vacuum breakers, mainly because vacuum interrupters can interrupt high-frequency currents — allowing the load-side voltage to be trapped again at a more severe value. More generally, it becomes relevant where the high-frequency current is within the interrupting capability of the interrupter; reignition transient frequencies below roughly 100 kHz are often cited as a range where the risk becomes more relevant, though this should be read as a guide rather than a hard boundary.

The grounding arrangement controls how much the reactor neutral can shift after the first pole clears, and that neutral shift changes both the voltage to ground on the reactor side and the recovery voltage across the breaker contacts. The standard treats three arrangements: the directly grounded case is the reference, and the other two are described as deviations from it. As a rule, the more the neutral is allowed to shift, the more severe the recovery-voltage duty becomes. The per-unit figures below are conditional on the assumptions in the guide’s worked example (in which the suppression peak overvoltage \(k_a\) is around 1.3 pu); they show the relative severity of the three arrangements rather than universal results, and the duty for a real installation should be calculated from the project circuit parameters.

For the ungrounded case, surge arresters to protect the reactor and the transformer tertiary are strongly advised. For the neutral-reactor case, if the neutral reactor remains in service during switching the recovery-voltage duty can be more severe and may not be represented by the standard test condition for a directly grounded reactor. Bypassing it before opening reduces the neutral shift and brings the switching duty closer to the directly grounded case, which is usually less severe and better represented by standard test conditions.

For shunt reactor switching the breaker is not selected mainly by the reactor current — that is easy to interrupt. It is selected by whether it can withstand the recovery voltage across the contacts without reignition.

How high both quantities go is set by a single parameter — the neutral-shift factor \(K\), the per-unit bias the reactor neutral takes up after the first pole clears: \(K=0\) directly grounded, \(0.5\) ungrounded, and an intermediate value fixed by the neutral-to-main inductance ratio for a neutral reactor. A larger \(K\) means a larger neutral shift, and therefore a larger voltage across the remaining open contacts. Writing the recovery peak to ground as \(k_c\) and the recovery voltage across the open contacts as \(k_{rv}\):

With \(\alpha=1\) this gives \(k_{rv}=1+k_a\) grounded, \(k_a+2\) ungrounded, and \(1+k_a+2K\) for the neutral-reactor case — the across-the-breaker figures used in Section 5. It also shows why the ungrounded and neutral arrangements are the more onerous: the neutral shift \(K\) adds directly to both the ground-referenced stress on the arrester and the across-contact stress that governs reignition.

Symbol note. On this page \(k_a\), \(k_c\) and \(k_{rv}\) denote the reactor-switching suppression peak, recovery-to-ground and across-contact overvoltages. Do not confuse \(k_c\) here with the capacitive-switching voltage factor \(k_c\) used on the capacitor-switching pages.

Reignition overvoltages, by contrast, depend strongly on the user’s own circuit and should normally be assessed by system calculation or EMT simulation rather than specified as a generic breaker rating.

The overvoltages cannot be eliminated, but they can be held to acceptable values by the following means. These do not all act in the same way: surge arresters limit the reactor terminal voltage to ground, metal-oxide varistors across the breaker limit the contact-gap voltage, opening resistors soften the interruption, and controlled opening lowers the probability of reignition by improving the contact-parting instant.

Two of these give a firm quantitative ceiling, because each holds a voltage to its own protective level \(k_{var}\) (in per unit of \(V_o\)). A varistor across the breaker — connected like a grading capacitor — holds the contact-gap voltage to \(k_{var}\), which caps the chopping and reignition overvoltages of Sections 2 and 3:

Choosing \(k_{var}\) therefore sets the switching-overvoltage ceiling directly. The difference between the two devices is where they act: a reactor-side arrester limits the voltage to ground at the reactor terminal, whereas a varistor across the breaker limits the voltage directly across the open contacts. Both conduct only on the fraction of the oscillation that exceeds their characteristic, so the energy they absorb is a small part of that stored in the reactor.

This is the part most often described loosely, so it is worth being precise. Controlled switching means the breaker operates at a chosen instant rather than at random, using a controller referenced to the source-side voltage (or the breaker current).

Controlled closing is usually less critical than controlled opening for shunt reactors, because reactor energisation does not produce the capacitor-bank-type inrush — the transient is 1.5 pu or less for a directly grounded reactor (1.73 pu for the second pole to make on an ungrounded reactor). It is not, however, irrelevant: for iron-core reactors, residual flux, saturation, asymmetry and slow-closing mechanisms can still make point-on-wave closing relevant, and a slow-closing device can cause repeated pre-striking and possible escalation.

Drawing the above together, the specification should define not only the reactor rating, but also the circuit capacitance, grounding arrangement, allowed suppression peak, dielectric withstand, and expected switching frequency. The standard asks the user to state:

Item 4 deserves care, because the load-side capacitance \(C_L\) is rarely just the reactor. It is the reactor together with the connection — busbar, cable, or GIS — and every item of plant between the breaker and the reactor: instrument transformers, CVTs, surge arresters, disconnectors, and bushings. The connection type dominates: a cable or GIS run contributes far more per metre than open busbar and can rival the reactor’s own capacitance, so it must be counted. The source-side capacitance, by contrast, is normally at least ten times larger and is not the limiting term. Indicative values, after IEEE C37.015 Annex B:

This example shows how a laboratory-measured chopping number is converted into a field overvoltage prediction for a specific reactor, breaker and grounding arrangement. The chopping number is an inherent property of the interrupter, largely independent of the test circuit, so it carries across to a different installation: measure \(\lambda = i_{ch}/\sqrt{C_t}\) on a test breaker, characterise it statistically, convert it into a 2% design value, then feed it into the field circuit to obtain \(k_a\), \(k_{rv}\), \(k_p\) and \(k_s\) (Sections 2, 3 and 6) and compare the mitigation options.

Because chopping performance varies from operation to operation, a statistical 2% value is used as the design figure — the level expected to be exceeded in only a small proportion (under 2%) of operations. Where the chopping number does not depend on arcing time it follows a normal distribution; where it does, a straight line is fitted to \(\lambda\) against arcing time by linear regression:

The guide works this through for a two-interrupter SF₆ breaker predicted on a 60 Hz, 500 kV system switching a directly grounded 525 kV, 135 Mvar reactor, comparing three options: no mitigation, a metal-oxide varistor across the breaker (\(k_{var}=1.6\) pu), and controlled opening.

The breaker’s wet switching-surge withstand across the open contacts is 3.15 pu; applying an 80% design margin gives a 2.52 pu ceiling. Option 1’s \(k_{rv}=2.31\) pu already clears it, so no mitigation is strictly essential for the breaker — but the reactor itself still swings to \(k_s=3.1\) pu, so its arrester (protective level ≈ 2 pu) will operate. The varistor and controlled-opening options cut the stress on both the breaker and the reactor, and controlled opening removes reignition altogether.

For the same reactor grounded through a neutral reactor the duty is markedly higher — the neutral shift increases the recovery voltage, giving \(k_{rv}\approx2.92\) pu and \(k_s\approx4.0\) pu without mitigation. Without bypassing or additional overvoltage limitation the same breaker duty may no longer be acceptable, and controlled opening qualifies only if the neutral reactor is bypassed before the breaker opens.

A reactor breaker has an easy current to interrupt but a demanding voltage to withstand. Current chopping launches a load-side oscillation; if the contact gap is not yet strong enough, the breaker reignites, producing steep transients that stress the reactor winding, and in the worst case these can escalate.

Grounding sets how severe the recovery voltage is: lowest when directly grounded, higher when ungrounded, and higher still when grounded through a neutral reactor (which is why that neutral reactor is usually bypassed before switching). Overvoltages are managed with surge arresters, opening resistors, varistors across the breaker, or — most effectively — controlled opening, where the contacts are parted at an instant that gives enough arcing time for a clean, reignition-free interruption.

The application message is simple: shunt reactor switching is a small-current but high-recovery-voltage duty. Focus on current chopping, recovery voltage, reignition probability, the grounding arrangement, and overvoltage mitigation.

The standard and further reading behind this guide.