Directional overcurrent protection (ANSI 67) — why an ordinary overcurrent relay falls short on a ring network or double-fed busbar

The guide on busbar differential protection (87B) and the guide on breaker failure protection (50BF) both cover protection functions in a network where the current direction through a feeder is more or less fixed: from the incomer toward the loads. This article covers what happens once that assumption no longer holds, and the protection function designed for that case: directional overcurrent protection, in practice denoted by the ANSI code 67 (and 67N for the earth-fault variant).

Why an ordinary overcurrent relay falls short with two possible current directions

An ordinary, non-directional overcurrent relay (ANSI 50/51) operates as soon as the current through its measuring point exceeds a set threshold, regardless of the direction in which that current flows. As long as a feeder can only be supplied one way, that is not a problem: any overcurrent in that feeder indicates a fault that the feeder itself must clear. As soon as a feeder can be fed from two directions, however — for example a ring network with two feeding points, a busbar with two parallel transformers, or an installation with distributed generation that can also supply fault current from the customer side during a fault — the same overcurrent threshold can be exceeded by a fault in either direction. An ordinary overcurrent relay cannot make that distinction and could operate on any fault, including one that should actually have been cleared by a different breaker — undermining selectivity.

Working principle: comparing current with a polarizing voltage

A directional relay (67) solves this by not only measuring the magnitude of the current, but also comparing its phase angle with a reference or polarizing quantity — typically a line voltage that remains reasonably stable and whose phase angle is known, even during a fault elsewhere in the network:

For a fault in the monitored direction, the phase angle of the fault current falls within a preset window around the so-called maximum torque angle (MTA) — the phase angle at which a fault current is expected for a typical fault in that direction — and the relay operates.
For a fault in the opposite direction, the phase angle of the current (relative to the same polarizing voltage) is rotated roughly 180° from the MTA, outside the preset window, and the relay stays quiet — even though the current exceeds exactly the same magnitude threshold as for a fault in the monitored direction.

For phase faults, a line voltage between two other phases is typically used as the polarizing quantity; for earth faults (67N), the neutral voltage or the residual current of a parallel current transformer is often used for polarization — the principle (comparing phase angle to a stable reference) remains the same in both cases.

Typical applications: where one direction is not enough

Ring networks: a cable ring fed from two points can, in the event of a fault somewhere in the ring, receive fault current from both directions; directional relays on both sides of each section ensure that only the breakers actually enclosing the failed section trip.
Parallel transformers on a shared busbar: for a fault on the low-voltage side of one transformer, the other, parallel-connected transformer can feed current back through the busbar; a directional relay on the feeding side of each transformer distinguishes its own fault from a back-feed caused by a fault elsewhere.
Installations with distributed generation (for example a large PV installation or CHP unit, see the guide on CHP grid connection): for a fault on the grid side, the generation itself can feed fault current back into the grid; directional protection at the connection point distinguishes this back-feed from a normal fault current coming from the grid.

Why the polarizing voltage must be chosen carefully

The reliability of a directional relay depends entirely on the polarizing quantity: if that voltage collapses or distorts precisely during the fault the relay is meant to detect (for example a three-phase fault close to the measuring point itself, where all three phase voltages collapse sharply), the phase-angle measurement can become unreliable at exactly the moment the protection is needed most. Modern numerical relays address this with techniques such as voltage memory polarization: the phase angle of the voltage just before the fault is briefly retained and used as a reference for as long as the actual voltage is too low to measure reliably.

Note: the exact MTA setting, the polarizing signal to use (line voltage, neutral voltage, or residual current) and the coordination with the non-directional overcurrent setting follow from the system study for the specific network; this article covers the principle, not a ready-made setting table for every network configuration.

Practical relevance

When assessing the protection concept of a ring network, a busbar with multiple incomers, or a connection point with distributed generation, it is important to check whether overcurrent protection that can be exposed to both directions has actually been implemented as directional — a non-directional overcurrent relay at such a point can undermine the selectivity of the whole switchgear without this coming to light during a normal, single-fed fault.

Common mistakes

Applying an ordinary, non-directional overcurrent relay at a point that can be fed from two directions — this can lead to nuisance tripping for a fault that should actually have been cleared by a different breaker.
Choosing the wrong polarizing voltage, or not checking that it has sufficient amplitude during a fault close to the measuring point — without reliable polarization (or voltage memory), the directional determination can become unreliable precisely during the most severe faults.
Not coordinating the MTA setting with the actual network impedance angle — an incorrectly set MTA can reduce the relay's sensitivity to the actual expected fault direction.
Forgetting directional protection when connecting new distributed generation to an existing network — without this adjustment, a back-feed during a grid fault can cause existing non-directional protection to operate unexpectedly, or fail to operate when it should.