Faults & Protection: When Things Go Wrong

The short circuit: when the grid's safety valve breaks

Everything the grid does is built on one quiet assumption: electricity flows where we tell it to, through the loads we connect — motors, lamps, heaters — each of which politely limits how much current it draws. A fault is what happens when that assumption shatters. A tree falls across a line, an insulator cracks in a storm, a cable's insulation ages and breaks down, or lightning ionises the air into a conducting path. Suddenly there is a near-perfect connection where there should be none — a wire touching the ground, or two phases touching each other. The load that was supposed to limit the current is simply gone, bypassed.

Think of the grid as a vast water network under enormous pressure. Normally every tap is a careful restriction, sipping a controlled trickle. A fault is a fire hose torn clean off its valve: the full pressure of the whole network now drives water through an opening with almost nothing holding it back. In electrical terms, the only thing left to limit the current is the small leftover impedance of the generators, transformers, and the line itself — and that is tiny. So the current explodes.

How big is a fault current, really?

Numbers make this visceral. A neighbourhood feeder might normally carry a few hundred amps. During a fault, the current at a major substation can leap to 20,000, 40,000, even 80,000 amps — for a few cycles, before anything trips. That is not a metaphor for danger; it is danger you can measure. At those levels the electromagnetic forces between adjacent busbars can bend solid copper. The heat (remember the I²R loss from earlier in this track?) can melt conductors and set insulation ablaze. An uncleared fault is how transformers explode and substations catch fire.

How do engineers compute that number before the fault ever happens? Here is where this guide quietly cashes in everything you learned earlier. The very same network model you built for load flow — buses, lines, transformers, all expressed in the per-unit system — is reused almost untouched for fault analysis. The genius of per-unit shines here: by Ohm's law, the fault current at a bus is roughly 1 ÷ Z, where Z is the total per-unit impedance from the source down to the fault point. A small impedance means a huge current. That's the whole story in one fraction.

Three-phase fault on a bus — back-of-envelope (per-unit)

Given  (all on a common 100 MVA base):
   Generator reactance ......... Xg = 0.20 pu
   Transformer reactance ....... Xt = 0.10 pu
   Line up to the fault ........ Xl = 0.10 pu
   Total source impedance ...... Z  = 0.40 pu

Fault current (per-unit) :
   I_fault = V / Z = 1.0 / 0.40 = 2.5 pu

Convert to real amps (say an 11 kV, 100 MVA base):
   I_base   = 100e6 / (sqrt(3) * 11e3) = 5,249 A
   I_fault  = 2.5 * 5,249           ~= 13,100 A

   --> a 'normal' load might be ~500 A here.
       The fault draws ~26x that, instantly.

The guardians: relays that see, breakers that act

A fault current that big cannot be allowed to flow for long — milliseconds matter. So the grid is patrolled by two kinds of device working as a team. The brain is the protection relay: a device that constantly watches the current (and often voltage) and decides, in a fraction of a cycle, whether what it's seeing is normal or a fault. The muscle is the circuit breaker: a heavy switch capable of interrupting tens of thousands of amps. The relay sees; the breaker acts. The relay can't break current, and the breaker can't think — together they are the immune system of the grid.

How does a relay *know* it's a fault? The classic answer is overcurrent: if the current blows past a set threshold, trip. But cleverer relays read more. A distance relay measures the ratio of voltage to current — which, by Ohm's law, is an impedance — and because impedance is proportional to the length of line up to the fault, the relay can literally estimate *how far away* the fault is, and trip only if it's within its zone. A differential relay is craftier still: it compares the current flowing *into* a protected zone (say, a transformer) with the current flowing *out*. In healthy operation they match exactly; if even a little current goes missing, it must be leaking out through a fault inside the zone — trip instantly.

DIFFERENTIAL PROTECTION  (guarding a transformer)

        I_in -->[ CT ]            [ CT ]--> I_out
                   \                /
                    \   relay     /
                     v  compares v
                    +--------------+
                    | I_in - I_out |  = I_diff
                    +--------------+
                            |
     HEALTHY:  I_diff ~ 0  -> everything that goes in comes out -> stay closed
     FAULT:    I_diff large -> current vanishing inside the zone -> TRIP breaker

CT = current transformer: scales 10,000s of amps down to a few amps
     the relay can safely measure.

Coordination: trip the right breaker, and only that one

Here is the subtle part that separates a good protection scheme from a clumsy one. A single fault deep in a neighbourhood is *seen* by many relays at once — the local one, the one upstream at the substation, the one further upstream still. If they all trip, the fault is cleared, yes, but so is power to half the city. That's a sledgehammer. What we want is selectivity: only the breaker *closest to the fault, on the source side* should open, isolating the smallest possible chunk of grid. Everyone else should hold their breath and stay closed.

How do you arrange that, when several relays all see the same surge? The classic trick is time grading. You give each relay a deliberate, increasing time delay as you move *upstream* toward the source. The relay nearest the fault is set to trip fastest. The next one up waits a little longer — a coordination time interval of perhaps 0.3 seconds — giving the downstream breaker a chance to clear the fault first. If it does, the upstream relay sees the current vanish and stands down. Only if the nearest breaker *fails* does the next one up step in as backup. It's a chain of guardians, each covering the one below it.

TIME-GRADED COORDINATION (radial feeder)

  SOURCE ===[CB-A]======[CB-B]======[CB-C]====== X  <- fault here
             relay A      relay B      relay C
             trip @0.9s   trip @0.6s   trip @0.3s
             (backup)     (backup)     (primary)

  t=0.00s : fault strikes; A, B, C ALL see the huge current
  t=0.30s : C times out first -> CB-C opens -> fault cleared
  t=0.31s : A and B see current vanish -> they reset, stay closed

  Result: only the small segment past CB-C goes dark.
          The rest of the feeder never blinks.

  If CB-C had FAILED to open:
  t=0.60s : B times out as backup -> CB-B opens -> fault cleared
            (bigger outage, but the grid is still saved)

The full sequence: from lightning to lights-back-on

Let's run the whole event in slow motion, because the real grid runs it faster than you can flinch. A lightning strike flashes over an insulator on a 132 kV line, opening a fault to ground. Watch the chain of guardians fire — and notice that, remarkably, most faults are temporary, so the grid even tries to heal itself before giving up.

1. t = 0 ms — the fault strikes. Lightning ionises the air across an insulator; a fault current of ~25,000 A erupts. On a three-phase line this may start on one phase and spread.
2. t ≈ 10–20 ms — the relay decides. The protection relay samples the current through its CT, confirms it's a genuine fault (not a momentary inrush), and issues a trip command — in well under one cycle.
3. t ≈ 40–60 ms — the breaker clears. The circuit breaker's contacts fly apart; an arc forms and is quenched (in SF₆ gas, oil, or a vacuum bottle). The faulted section is now dead and isolated. Total clearing time: often 2–3 cycles.
4. t ≈ 0.5–1 s — auto-reclose tries. Since the lightning's arc has long since cooled and de-ionised, an automatic recloser re-energises the line. If the fault was temporary (it usually was), power is restored and customers barely noticed a flicker.
5. If the fault persists — a tree still resting on the line — the relay trips again and locks out. Now a human crew is dispatched. Better one locked-out line than a fire or a cascading blackout.

Why this rung matters — and where it leads

Step back and see what you've assembled. You can now picture a fault as a torn-off fire hose, estimate its current from the per-unit impedance with a single division, and follow the relay-and-breaker reflex arc that kills it in milliseconds while keeping the lights on everywhere else. Crucially, you saw that [[ee-fault-analysis|fault analysis]] is not a new subject — it's your load-flow network model, expressed in per-unit, asked a different question: not 'how does power flow normally?' but 'how much current floods in when something breaks?'

From here the path forks toward the frontier. Modern grids increasingly run on inverter-based renewables — solar and wind — which behave nothing like a spinning synchronous machine during a fault: they don't deliver a huge slug of short-circuit current, so the overcurrent relays we relied on for a century can be blinded. Rewriting protection for a grid with little spinning inertia is one of the live research problems of the energy transition. The guardians, in other words, are having to learn new instincts — and the integration of renewables is exactly why.