Substation Layout, Earthing and Insulation Coordination

Every diagram we have drawn since Chapter 3 has treated a bus as a dimensionless point. Chapter 16 gave that point an admittance, Chapter 18 gave it a specification, Chapter 25 gave it a fault level, and Chapter 35 gave it a circuit breaker. On the ground, that point is a fenced rectangle of land carrying steel structures, insulator stacks, current and voltage transformers, disconnectors, breakers, transformers, a control building and a buried copper mesh. The substation is where the abstraction is cashed in.

Four jobs are performed there, and a substation is usually classified by which of them dominates. It transforms voltage — a step-up substation at a generating station lifts \(11\)–\(21\;\mathrm{kV}\) to \(220\) or \(400\;\mathrm{kV}\); a grid substation steps \(400\) down to \(220\), a secondary one \(220\) to \(132\) or \(66\), and a distribution substation \(33\;\mathrm{kV}\) down to \(11\;\mathrm{kV}\) and finally to \(415\;\mathrm{V}\). It switches — a switching substation contains no transformer at all and exists purely so that lines can be connected, separated and isolated. It measures and protects — every instrument transformer feeding the relays of Chapter 36 lives here. And it controls reactive power — the capacitor banks, shunt reactors and static compensators of Chapter 34 are installed on substation buses because that is where the network is accessible.

A fifth category, the converter substation, terminates an HVDC link; Chapter 38 is devoted to what happens inside it.

Physically the equipment falls into a short list. Busbars — rigid aluminium tubes or strung ACSR — are the common connection points. Circuit breakers interrupt load and fault current; disconnectors (isolators) carry no interrupting duty at all and are operated only on a dead circuit, to establish a visible open point for safety. Earthing switches bond an isolated section to the grid before anybody touches it. Current and voltage transformers scale the measurands. Surge arresters cap the overvoltage. Wave traps block carrier signals from leaving the protected zone. Above everything run the earth wires that shield the yard from direct lightning strokes, and beneath everything lies the earthing grid of Section 37-6.

The design question is narrow enough to be answered systematically. Given \(n\) circuits — lines and transformers — that must meet at a station, how many buses and how many breakers do we install, and how are they interconnected? Every scheme in use is one of six answers, and they form an almost perfect ladder of increasing cost and increasing security.

Single bus. One busbar; each circuit taps it through one breaker flanked by two disconnectors. It is the cheapest arrangement possible — one breaker per circuit — and the least secure. A fault on the bus, or on any breaker, trips the entire station. Maintenance of a breaker requires its circuit to be de-energised. It is used at distribution voltages and in small industrial stations where a total outage is tolerable.

Sectionalised single bus. Insert a bus-coupler breaker in the middle of the bus, and feed alternate circuits from the two halves. A bus fault now costs half the station instead of all of it, and half the station can be taken out for maintenance while the other half runs. The cost is one extra breaker for the whole station — an outstanding return, which is why almost no genuinely single bus is built above \(11\;\mathrm{kV}\).

Main and transfer bus. A second bus, the transfer bus, runs parallel to the main bus, and a single bus-coupler breaker joins the two. Every circuit can be switched onto the transfer bus and picked up by the coupler breaker, freeing its own breaker for maintenance while the circuit stays in service. The scheme therefore solves the breaker maintenance problem with one extra breaker, but not the bus fault problem: a main-bus fault still trips everything. During a transfer, the circuit is protected by the coupler's relays rather than its own, which complicates the settings of Chapter 36.

Double bus, single breaker. Two full-capacity buses, each circuit selectable onto either by a pair of disconnectors, with a coupler breaker between them. Circuits may be split into two groups, and a bus fault costs only the group on that bus, since the healthy group is switched across. Flexibility is high; the breaker count remains \(n+1\); but a breaker fault still removes its circuit, and the changeover involves disconnector operations.

Double bus, double breaker. Each circuit connects to both buses through its own breaker: two breakers per circuit. Any bus can be lost, any breaker can be lost, and every circuit stays in service, switched by the other breaker. It is the most secure arrangement known and, at \(2n\) breakers, easily the most expensive. It appears only where the load is critical and the voltage is high enough that an outage is intolerable.

Ring bus. Close the bus on itself so that the breakers form a ring and each circuit is tapped between two adjacent breakers. There are \(n\) breakers for \(n\) circuits — no more than a single bus — yet every circuit is fed from two directions. A fault on one circuit opens the two breakers flanking it and the ring becomes an open chain; all other circuits stay energised. The weakness is what happens next: while the ring is open, a second fault splits the station, and a breaker taken out for maintenance opens the ring in the same way. Rings are therefore restricted in practice to about four to six circuits.

Breaker-and-a-half. Between the two buses, string three breakers in series to form a diameter; tap one circuit between the first and second breakers and another circuit between the second and third. Two circuits share three breakers, so the count is \(1.5\) breakers per circuit — the name. The centre breaker is shared. A bus fault trips only the breakers on that bus and every circuit continues to be fed from the other bus through its centre breaker. Any breaker may be removed for maintenance without disconnecting a circuit. It gives very nearly the security of double-bus-double-breaker at three-quarters of the breaker count, and it is the standard arrangement of large \(400\;\mathrm{kV}\) and \(765\;\mathrm{kV}\) substations worldwide.

Everything above assumed the insulating medium is atmospheric air, which breaks down at roughly \(30\;\mathrm{kV/cm}\) in a uniform field and considerably less in the divergent fields around real hardware. Clearances are therefore large: a \(400\;\mathrm{kV}\) air-insulated yard needs phase-to-earth clearances of about \(3.5\;\mathrm{m}\) and occupies several hectares. In a city centre that land does not exist, or costs more than the whole substation.

Sulphur hexafluoride, \(\mathrm{SF_6}\), changes the arithmetic. It is an electronegative gas: its molecules capture free electrons and remove them from the avalanche that would otherwise develop into a breakdown. At atmospheric pressure its dielectric strength is about \(2.5\) to \(3\) times that of air, and because strength continues to rise with pressure, at the \(3\)–\(5\;\mathrm{bar}\) used in practice it exceeds air by roughly an order of magnitude. The same gas is an excellent arc-quenching medium, which is why Chapter 35 gave \(\mathrm{SF_6}\) breakers their own section.

In a gas-insulated substation the busbars, disconnectors, earthing switches, current and voltage transformers and the breakers themselves are enclosed in earthed aluminium tubes filled with pressurised \(\mathrm{SF_6}\). The live conductor runs along the axis, supported on cast-resin spacers that also divide the enclosure into separately monitored gas compartments. The result is a coaxial system: fields are almost uniform, so the clearance can be cut to the theoretical minimum, and the whole \(400\;\mathrm{kV}\) station shrinks to perhaps \(10\%\) of the equivalent air-insulated footprint.

Chapter 23 built the zero-sequence network and Chapter 24 showed that the single line-to-ground fault current is

The impedance \(Z_n\) deliberately inserted between the transformer or generator star point and earth appears tripled in the zero-sequence path and nowhere else. That single algebraic fact is the entire theory of neutral earthing: it is the one knob that changes earth-fault current without touching anything else in the network.

Why not simply leave the neutral unearthed? An isolated neutral system does have one attraction — a single ground fault draws only the small capacitive charging current of the healthy phases, so the system can keep running with the fault on it. But the price is severe. The neutral point shifts by the full phase voltage, so the two healthy phases are stressed to \(\sqrt3\) times normal for as long as the fault lasts, and every piece of insulation must be rated accordingly. Worse, the fault current is small enough to extinguish at a current zero and restrike on the next voltage peak, trapping charge on the line capacitance and escalating in steps — an arcing ground, which can build transient overvoltages to \(5\)–\(6\) times normal and destroy insulation elsewhere in the system.

Solid earthing sets \(Z_n = 0\). The earth-fault factor falls to about \(1.4\) or below, insulation can be graded — the winding near the neutral of a large transformer is wound with less insulation than the line end, a real saving — and the fault current is large, which makes it easy for the earth-fault relays of Chapter 36 to see. The cost is exactly that large current: the single line-to-ground fault current can exceed the three-phase value when \(Z_0 < Z_1\), setting the breaker duty of Chapter 25 and imposing severe mechanical and thermal stress. Solid earthing is universal at \(66\;\mathrm{kV}\) and above.

Resistance earthing inserts \(R_n\). It limits the fault current, and because \(R_n\) is resistive it also damps the transient oscillation between the line capacitance and the system inductance, which suppresses arcing-ground escalation. The current is normally limited to somewhere between full-load current and about \(400\;\mathrm{A}\); enough for selective relaying, small enough that the damage at the fault point is limited. The resistor dissipates \(I_f^2 R_n\), which for a few hundred amperes and tens of ohms is megawatts — it is rated for a short time only, typically \(10\;\mathrm{s}\), consistent with the fault clearing time. Generator neutrals are almost always resistance-earthed, often through a distribution transformer with a low-ohmic secondary resistor, to keep stator iron damage below the threshold at which a rewind becomes a re-core.

Reactance earthing inserts \(X_n\). It limits the current with negligible loss, and it is the usual way to trim a generator or transformer neutral so that the single line-to-ground current does not exceed the three-phase current. The reactance must be kept small — \(X_0 \le 3X_1\) overall — because a large \(X_n\) pushes the system toward resonance with the line capacitance and toward the very restriking problem earthing was meant to prevent. That resonance, however, is not always an enemy. Deliberately tuned, it becomes the Peterson coil.

Consider a three-phase line with capacitance \(C\) from each phase to earth, and let phase \(a\) develop a solid ground fault. Phase \(a\) is now at earth potential, so its own capacitance is short-circuited and carries nothing. The two healthy phases, however, are raised to the full line voltage with respect to earth: \(V_b\) and \(V_c\) become \(\sqrt3\,V_{ph}\) in magnitude, \(60^\circ\) apart in phase, and each drives a charging current through its own capacitance.

The capacitive current arriving at the fault is therefore \(3\omega C V_{ph}\) — three times what a single phase would contribute, which is the same factor of three that appears in the zero-sequence algebra of Chapter 22, and for the same reason: all three phase-to-earth capacitances are in parallel as seen from the neutral.

Waldemar Petersen's idea, published in 1916, was to connect an inductor between the neutral and earth so that its current at the fault is equal and opposite. The coil sits at phase-voltage \(V_{ph}\) across it once the neutral has shifted, so it draws

The coil current lags the neutral voltage by \(90^\circ\) and the capacitive current leads it by \(90^\circ\); they are in exact antiphase at the fault point. When the tuning condition holds, the current in the arc is reduced to the small residual made up of the coil's copper loss, the line's leakage conductance and the harmonic content — typically a few percent of \(I_C\). That residual is far too small to sustain an arc, so the fault self-extinguishes and the system continues to run with one phase earthed.

Two practical points follow immediately. First, \(C\) depends on how much line is connected, so a coil tuned for the full network is detuned the moment a feeder is switched out. Real arc-suppression coils are therefore built with a movable plunger or a tapped winding and an automatic controller that measures the neutral displacement voltage and re-tunes continuously; the residual is held to a few percent of \(I_C\). Second, resonant earthing is only worthwhile where the charging current is modest — up to a few hundred amperes — which in practice means \(33\;\mathrm{kV}\) and below on predominantly overhead networks. Cable systems have capacitances one to two orders of magnitude larger (Chapter 9), and the coil would be impossibly big.

Everything so far concerned the neutral connection at a single point. The other earthing problem is different in kind: when fault current flows into the soil, the soil is not an equipotential, and people standing in the yard are in the resulting field.

Bury a horizontal mesh of copper conductors — typically \(7\;\mathrm{m}\) to \(15\;\mathrm{m}\) spacing at \(0.5\;\mathrm{m}\) depth, supplemented by driven rods — and bond every structure, fence, enclosure and neutral to it. When a ground fault injects current \(I_G\) into that grid, the whole grid rises to a potential above remote earth called the ground potential rise:

For a grid of buried length \(L\) covering area \(A\) in soil of resistivity \(\rho\), the resistance to remote earth follows Laurent and Niemann's approximation — a hemispherical-plate term plus a conductor-length term:

The first term dominates for a large grid: resistance falls as \(1/\sqrt A\), so doubling the area buys only a \(30\%\) reduction. Adding more conductor inside a fixed area helps the second term only, and quickly saturates. This is the fundamental frustration of earthing design — you cannot simply drive \(R_g\) to zero, and a \(1\;\Omega\) grid carrying \(20\;\mathrm{kA}\) is at \(20\;\mathrm{kV}\) above remote earth.

Since the GPR cannot be eliminated, safety is defined not by the grid's absolute potential but by the differences a human body can be exposed to.

What matters physiologically is the current through the heart, not the voltage. Dalziel's work fixes the tolerable body current for a shock of duration \(t_s\) seconds as \(I_B = k/\sqrt{t_s}\), with \(k = 0.116\) for a \(50\;\mathrm{kg}\) person and \(0.157\) for a \(70\;\mathrm{kg}\) person. Taking the body resistance as \(1000\;\Omega\) and each foot as a \(0.08\;\mathrm{m}\)-radius plate of resistance \(3\rho_s\), the two feet appear in parallel for touch (\(1.5\rho_s\)) and in series for step (\(6\rho_s\)). Multiplying the tolerable current by the total circuit resistance gives the tolerable voltages.

Here \(\rho_s\) is the resistivity of a surface layer of thickness \(h_s\) — usually \(0.10\)–\(0.15\;\mathrm{m}\) of crushed rock, at \(2000\)–\(5000\;\Omega\mathrm{\cdot m}\) when wet — and \(C_s\) is the derating factor that accounts for the layer being thin enough that the foot still "sees" the native soil beneath. The crushed rock is not decoration: as Example 3 shows, it multiplies the tolerable touch voltage by roughly three, and it is the cheapest safety measure available in a substation yard.

The actual voltages the grid produces are the mesh voltage \(E_m\) — the worst touch voltage, found at the centre of a corner mesh where the surface potential dips furthest below the grid — and the step voltage \(E_s\), largest just outside the perimeter. IEEE 80 gives them as

with \(K_m\) and \(K_s\) geometric spacing factors, \(K_i\) an irregularity factor, and \(L_M, L_S\) effective buried lengths. The design is safe when \(E_m \le E_{\text{touch}}\) and \(E_s \le E_{\text{step}}\). If it is not, the remedies are ordered by cost: add crushed rock, reduce the mesh spacing, add ground rods around the perimeter, extend the grid area, and — last — reduce \(I_G\) by changing the earthing method of Section 37-4.

The insulation of a substation is never designed for the working voltage. It is designed for the worst transient that will reach it, and those transients come in three families distinguished by their duration.

Temporary overvoltages last cycles to seconds and are at or near power frequency: the healthy-phase rise during an earth fault (Section 37-4), the Ferranti rise on a lightly loaded long line (Chapter 14), load rejection, and ferroresonance. They are modest in amplitude — \(1.2\) to \(1.5\;\mathrm{pu}\) — but they last long enough to matter for arrester energy, and they set the arrester's continuous operating voltage.

Switching overvoltages last hundreds of microseconds. Their origin is the travelling-wave behaviour of Chapter 12. Energising an unloaded line launches a step into a line whose far end is effectively open; the reflection coefficient there is \(+1\), so the wave doubles, and if the line already carries trapped charge from a previous opening — as it does on a rapid autoreclose — the step is measured from that trapped voltage and the arriving peak can approach \(3\;\mathrm{pu}\). Current chopping in the interruption of small inductive currents, discussed in Chapter 35, is another source. The standard test waveshape is \(250/2500\;\mu\mathrm{s}\).

Lightning overvoltages last microseconds. A return stroke injects a current of typically \(10\)–\(50\;\mathrm{kA}\) (with a median near \(30\;\mathrm{kA}\)) with a front of about \(1\)–\(2\;\mu\mathrm{s}\). If it strikes a phase conductor directly, the current divides in the two directions and produces a travelling wave of amplitude \(\tfrac12 I Z_0\); with \(Z_0 \approx 400\;\Omega\) even a modest \(10\;\mathrm{kA}\) stroke gives \(2\;\mathrm{MV}\), far above any insulation. The standard waveshape is \(1.2/50\;\mu\mathrm{s}\).

Shielding transfers the problem rather than removing it. The stroke current now flows down the tower and out through its footing resistance \(R_f\), lifting the tower top to \(I R_f\) plus the inductive drop \(L\,\mathrm{d}i/\mathrm{d}t\) along the tower steel. If that potential exceeds the insulator string's critical flashover voltage, the string flashes over from the earthed tower to the live conductor — a back-flashover. The remedy is to reduce \(R_f\): driven rods where the soil allows, or a buried counterpoise wire radiating from each footing where it does not. Example 6 works this through.

The overvoltage that finally reaches a transformer inside the substation is therefore not the stroke voltage but a travelling wave that has already been limited by shielding, by back-flashover, by corona attenuation on the span (Chapter 15) and, last of all, by the surge arrester.

A surge arrester is a device that is an insulator at power frequency and a conductor at surge voltage. Placed from line to earth beside the equipment it protects, it diverts the surge current to the earthing grid and clamps the terminal voltage to a value the insulation can withstand.

The crudest form is a rod gap: two electrodes separated by air. It is cheap and indestructible, but its breakdown voltage scatters widely, it does not clear the power-frequency follow current — the arc, once struck, is fed by the system until a breaker trips — and its volt–time characteristic rises steeply for fast fronts, so it may not operate before the protected insulation does. It survives only as a backup on distribution lines.

The valve-type arrester improved on this by putting a stack of series spark gaps in series with discs of silicon carbide. The gaps hold off the power-frequency voltage; when they break down, the SiC discs conduct, and because SiC is non-linear (\(I \propto V^{\,\alpha}\) with \(\alpha \approx 4\)–\(6\)) their resistance falls sharply during the surge and rises again afterwards, allowing the gaps to interrupt the follow current at the first current zero.

Modern practice is the gapless metal-oxide arrester. Zinc-oxide discs doped with bismuth, antimony and cobalt oxides have an exponent \(\alpha\) of \(25\) to \(50\): the resistance changes by six or more orders of magnitude for a factor-of-two change in voltage. That non-linearity is so extreme that no series gap is needed at all. At continuous operating voltage the arrester draws a leakage current of well under a milliampere, mostly capacitive; at a \(10\;\mathrm{kA}\) surge it conducts freely, and when the surge passes it returns to its insulating state on its own — there is no follow current to interrupt.

Selection starts from the temporary overvoltage, because that is what the arrester must survive without conducting. On an effectively earthed system the earth-fault factor is about \(1.4\) (Section 37-4), so with \(U_m\) the highest system voltage,

A non-effectively earthed system therefore needs an arrester rated roughly \(25\%\) higher, and since \(U_{res}\) scales with \(U_r\), it also gets a worse protective level for the same insulation. The earthing decision of Section 37-4 has come back, exactly as promised.

Placement is as important as rating. The arrester should stand as close to the protected apparatus as the layout allows, because two effects raise the voltage at the equipment above \(U_{res}\). The connecting leads carry the discharge current at high \(\mathrm{d}i/\mathrm{d}t\) and contribute \(L\,\mathrm{d}i/\mathrm{d}t\) at roughly \(1\;\mu\mathrm{H}\) per metre. And the separation distance \(S\) between arrester and transformer allows the incoming wave to keep rising for the round-trip travel time before the arrester's clamping action arrives back:

with \(v \approx 300\;\mathrm{m/\mu s}\) on an overhead line. A steepness of \(1000\;\mathrm{kV/\mu s}\) and a separation of \(20\;\mathrm{m}\) add \(133\;\mathrm{kV}\) — significant at \(220\;\mathrm{kV}\) and decisive at \(66\;\mathrm{kV}\).

Insulation coordination is the process of choosing the insulation strength of every item in a substation, and the protective level of the arresters, so that the two are correctly ordered: the arrester always operates first, and whatever flashover does occur happens where it does least harm.

The insulation strength of an item is stated as its Basic Insulation Level — strictly, the basic lightning impulse insulation level, BIL — which is the crest value of a standard \(1.2/50\;\mu\mathrm{s}\) impulse the item withstands. Above \(300\;\mathrm{kV}\) a second number, the switching impulse level (BSL, tested at \(250/2500\;\mu\mathrm{s}\)), becomes the governing figure, because long air gaps are relatively weaker against slow fronts than against fast ones.

The coordination itself reduces to one inequality. Let \(V_p\) be the protective level actually appearing at the equipment terminals — the \(V_{\text{eq}}\) of Section 37-8, not the bare \(U_{res}\). Then define

Standardised insulation levels are tabulated against the highest voltage for equipment \(U_m\), and normally two BIL values are offered at each \(U_m\) — a higher one for exposed installations and a lower one where arrester protection is close and the station is well shielded. Choosing the lower level is a legitimate economy, but only after the margin has been checked.

Take \(f = 50\;\mathrm{Hz}\) throughout, and quote the standard formulae you use. Problems 5 to 8 require the results of more than one section.