Directional overcurrent
The key idea
On a radial feeder the current has only one path, so its size alone is enough to protect the line. Close a ring or add a second source, and the same relay sees nearly the same current for a fault on either side of it. Only the angle tells the relay which side, and it measures that angle against a voltage.
The idea
On a radial feeder this problem never comes up. Power leaves the source and travels outward, so any current above the setting means a fault somewhere ahead. An overcurrent relay that measures only amps has everything it needs.
Close that feeder into a ring, or connect a second source at the far end, and the picture breaks. Current can now reach a fault from either direction. A relay in the middle of the line sees current for a fault on the section it protects, and also for a fault behind it, on the neighboring section. The two currents are often within a few percent of each other in size.
So the magnitude cannot name the faulted section. The angle can. Fault current is set almost entirely by reactance (a faulted line is inductive, with comparatively little resistance), so the current lags the voltage driving it by close to 90°. Give the relay a voltage to measure that lag against — the polarizing voltage — and the current's angle becomes a usable fact.
Now move the fault to the other side of the relay. The current in that conductor does not shrink or grow. It physically turns around — a full 180° on the phasor diagram.
A directional element exploits this by drawing a straight line through the origin of the phasor plane and testing which side the current falls on. The line's orientation comes from the maximum torque angle (MTA): the current angle the element is most sensitive to, set among the angles real faults actually produce. Current on the forward side of the line is allowed to trip; everything else is blocked.
Direction does not replace the overcurrent element; it only gates it. The relay still needs enough current, and it still waits its graded time. The directional check decides just one thing: whether this fault is this relay's business at all.
Try it
Slide the fault across the relay. The current barely changes in size — but the verdict flips completely.
through the relay: 2.50 pu at −90° vs V
FORWARD — relay may trip
The model uses two 1∠0 pu sources and pure reactance: 0.15 pu of source, 0.2 pu of line, and 0.15 pu of source. The relay sits 30% of the way along. A fault at 0.80 gives the relay 2.50 pu. A fault at 0.10 gives it 2.22 pu. The two currents are 11% apart in size, and they have opposite meanings. A ring needs direction for exactly that reason.
Why it matters
- A closed loop makes direction necessary. A radial feeder never needs it. But once a ring closes, or a generator connects at the far end, every relay inside the loop must know which direction it protects.
- Grading assumes an order that exists in one direction only. A graded set of curves runs from the fault back to the source. When current can arrive from either end, you grade two sets, one per direction, and the directional element picks the set that applies.
- A blocked relay is still working correctly. The relay nearest a fault on the neighboring section still measures thousands of amps. Without direction it trips, disconnecting a healthy section along with the faulted one, and the outage spreads past the zone that should have contained it.
- The reference is the weak point. A directional element lives on its voltage measurement. A blown VT fuse removes the reference and the element can no longer decide. That is why relays carry fuse-failure supervision, and why engineers choose a polarizing voltage the fault cannot collapse.
The math, if you want itOptional — the page reads completely without it
The element compares the current's angle against its characteristic and operates on the forward half of the plane:
directional operate criterion
cos( θI − MTA ) > 0
The relay measures θI from the polarizing voltage, and the MTA is the angle of highest sensitivity. The boundary sits 90° to each side of the MTA, so the operate region is exactly half the plane. The widget uses an MTA of −45°.
The current lands near −90°, because the reactance of a faulted line dominates its resistance:
fault current angle
θI = −arctan( X / R ) → −90° when X ≫ R
The reversal is a full 180°, and it is not an approximation. For a fault beyond the relay, the relay carries the left source's contribution flowing right; for a fault behind it, the right source's contribution flowing left. Same conductor, same measuring point; only the sign changes:
the two cases at the relay
θI = −90° forward · θI = +90° reverse
Two simplifications are worth naming: the widget models the network as pure reactance and polarizes against a single voltage reference, so every current comes out at exactly ±90°.
Real relays polarize a phase element from a quadrature voltage, the line-to-line voltage of the other two phases, because a close-in fault collapses the faulted phase's own voltage and would leave the element with no usable reference. Real networks also have resistance, so the fault angle stops short of 90°. That is why the MTA is set near the real fault angle rather than at −90° exactly.
See it in Phasor
Phasor knows the topology, so it knows which relays sit inside a loop: those are the overcurrent elements that need directional supervision. Set the direction and the MTA on each one, and the fault study reports the current every relay sees, with its angle, for a fault on either side of it.