Overcurrent protection on distribution feeders

A distribution feeder is protected by devices in series: a breaker at the substation, one or more reclosers down the trunk, fuses on the laterals. Every one of them sees the same fault. Protection engineering is mostly the work of making sure that only the right one operates, and that it operates before anything melts.

That reduces to two decisions, and everything else — pickup, curve shape, time dial, coordination interval — is machinery in service of them:

1. Is this a fault or is it load?
2. Is this fault mine to clear, or does something downstream own it?

Radial feeders make both decisions tractable. There is one source, so fault current flows one way and falls off with distance from the substation. Current magnitude alone carries enough information about how far away a fault is that you can protect the whole feeder with overcurrent elements and never measure impedance. That single assumption is what the rest of this rests on — and it is also the first thing distributed generation takes away.

The devices you actually set

Function numbers come from IEEE C37.2. On a feeder you will keep meeting the same handful:

The ground element deserves the emphasis. Phase elements must sit above load current, but the residual element measures 3I₀, which is near zero on a balanced feeder. That lets you set ground pickup at a fraction of phase pickup and catch high-impedance faults — a conductor lying on dry soil — that no phase element would ever see. Feeder imbalance and single-phase lateral loading set the practical floor, not the relay.

Pickup: a window, not a number

Phase pickup has to fit between two moving limits:

I_load,max  ×  margin   <   I_pickup   <   I_fault,min  /  reliability

The lower bound is not nameplate load. Two effects push it up:

• Transformer inrush, several times rated current for a few cycles.
• Cold load pickup, which is the harder one. After an extended outage the natural diversity of thermostats, motors and compressors is gone, and they all restart together. The feeder can draw multiples of its pre-outage load for seconds to minutes — long enough to sit inside an inverse-time curve rather than being ridden through.

The upper bound is the minimum fault current in the protected zone, which usually means a phase-to-phase or high-impedance ground fault at the far end, under minimum generation. On a long rural feeder those two bounds converge, and on a bad one the window closes entirely: the current drawn by a fault at the end of the line is no longer clearly distinguishable from a heavy cold-load restart. That is the point at which overcurrent alone stops being sufficient and the answer becomes voltage-restrained elements, a mid-line recloser to shorten the zone, or impedance-based protection.

The curve

Inverse-time characteristics exist because current magnitude is a usable proxy for electrical distance. One curve gives you two things at once: a close-in fault clears fast because it draws more current, and a downstream device with the same shape sits naturally below yours across the whole current range.

Two standards define the shapes, and they are not interchangeable.

IEC 60255-151:

        TMS · k
t = ─────────────────
    (I / Is)^α  −  1

IEEE C37.112:

     TD   ⎛       A              ⎞
t = ──── ·⎜ ───────────────  +  B⎟
      7   ⎝ (I / Is)^p − 1       ⎠

An IEC very-inverse curve and an IEEE very-inverse curve share a name and are different curves. Loading one into a relay when the coordination study assumed the other is a quiet, common and entirely avoidable error — the settings look right in the file and the margins are wrong on the feeder.

Choosing a shape is mostly about what you have to coordinate against:

• Extremely inverse tracks fuse melting characteristics most closely, so it is the usual choice on feeders with fused laterals.
• Standard / moderately inverse stays dependable across a wide range of fault currents and tolerates uncertainty in the short-circuit model.
• Definite time is the honest option where the current-versus-distance relationship is too flat for an inverse curve to discriminate anything.

Coordination is checked at the worst current, not a convenient one

Between two devices in series you need a coordination time interval — commonly 0.2–0.4 s. It is not an arbitrary safety factor; it is a sum:

• Breaker interrupting time, roughly 3–5 cycles, so 50–83 ms at 60 Hz.
• Relay overtravel, historically about 0.1 s on electromechanical relays from disc inertia, effectively zero on numerical relays.
• Characteristic and instrument tolerance — the standards define accuracy classes, and CT saturation at high current degrades them further.
• Margin.

Which is why 0.3 s is inherited practice and 0.2 s is defensible on a fully numerical scheme, while neither number means anything without the arithmetic behind it.

The mistake worth naming: verifying coordination at a single current. Two inverse curves converge as current rises, so a pair that looks comfortably separated at minimum fault current can cross at maximum fault current — right where the fault energy is highest. Coordination has to hold across the whole range of currents both devices can see.

Fuse saving versus fuse clearing

Most distribution faults — commonly cited as the large majority — are transient: a branch, an animal, a flashover that does not survive a de-energized interval. Reclosing exists to exploit that, and it forces a policy choice on every fused lateral.

Fuse saving puts a fast curve on the recloser so it operates before the lateral fuse melts, then recloses. A transient fault on a lateral costs nobody a fuse and nobody a truck. The price is that every customer on the feeder sees a momentary interruption for a fault on one lateral.

Fuse clearing lets the fuse blow. Only the lateral is lost, the rest of the feeder never blinks — but the outage is sustained until a crew arrives.

The trade has shifted. Momentary interruptions are far more expensive to customers with electronics and variable-speed drives than they were when fuse saving became standard practice, which has moved many utilities toward fuse clearing on laterals.

What distributed generation breaks

Everything above assumes one source and unidirectional flow. Add generation downstream and specific assumptions fail:

• Direction stops implying location. A feeder with generation can push fault current toward the substation, which is what 67 is for.
• Infeed causes underreach. For a fault beyond a DER connection point, the substation relay sees only part of the total fault current — the DER supplies the rest. The relay reads a smaller current than the fault actually is, and its inverse curve therefore trips slower than the study predicted.
• Reclosing can close out of phase. If DER keeps an island energized through the dead time, the reclose is a closure between two unsynchronized sources. IEEE 1547 requires DER to cease to energize on an abnormal condition, but the timing has to be checked against the actual reclose interval rather than assumed.
• Sympathetic tripping. Generation on a healthy feeder can contribute enough current to a fault on an adjacent one to operate its own protection.

Standards worth keeping open

The thread running through all of it: coordination is a property of the system, not a setting on a relay. A device whose settings are individually defensible can still be wrong, because correctness is defined by what the device upstream and the device downstream are set to do at the same current, at the same instant.