Textbooks say about 80% of faults are temporary. Breaker counts agree: 89% of breaker operations reclose and hold. But across 64,382 fault sequences on 676 US feeders, load was lost on 36% to 53% of reclose attempts.
It started as an assertion, not a count. These quotations trace the figure from breaker openings to faults.
"More than 50 per cent of the short circuits on these feeders are of a temporary nature, and therefore clear themselves as soon as the feeder trips out" (p. 645)1
The earliest share of temporary faults located, from operating experience; no count given.
"from 75 to 90 per cent of the breaker openings are caused by temporary faults which disappear after the first, second or third reclosure" (p. 737)234
No data; repeated in 1934. Earliest share of breaker operations located: Blake (GE), 1925, 58% held on the first reclose.
"experience shows that from 75 to 90 per cent of the faults are momentary and self-clearing, but will blow a fuse" (p. 25)56
The same numbers, now said of faults, including faults beyond fuses that no breaker counted. Pacific Electric repeats the sentence in 1937.
"According to reliable statistics, approximately 85 per cent of automatic interruptions ... result from transient disturbances" (p. 1012)7
Its own table counts breaker trips: 83% held on the first reclose. The T&D Reference Book later reprints it.
"approximately 85 per cent" (Kyle, 1945, p. 261); "about 95%" (REA, 1945, p. 26); "about 80 per cent" (Westinghouse, 1947, p. 256)8910
Two recloser makers and the REA's sectionalizing guide. None of the three gives data.
Initially, 75 - 90% was asserted of breaker openings; by 1934 it was said of faults, fuses included.
Round, made-up numbers on a simple feeder: enough to show what a breaker count measures, and what it misses.
On a lateral: 40 faults
Fast trip, reclose: the fault is gone and the fuse is saved.
On the mainline: 10 faults
Fast trip, reclose: the fault is gone.
On a lateral: 40 faults
Fast trip, reclose; the fault returns and the fuse blows. The breaker holds.
On the mainline: 10 faults
Fast trip, reclose; the fault returns. Slow trip: lockout.
Outcome 3 is the gap. To a breaker count it looks exactly like outcome 1: one fast trip, one reclose that held. But the fault was still there, the fuse blew, and the lateral is out.
Count only the breaker
Customers without power
Looking only at the substation breaker gives a 90% success rate.
But 40 of its 90 "successes" (44%) left a lateral without power, because the fault was permanent.
A breaker that held 90% of the time does not mean 90% of faults are temporary.
A breaker that recloses and holds says nothing about the fuse that blew behind it. So we followed the load. Of 46,726 reclose attempts, 47% were likely successes: nothing lost, or the lost load back within minutes.
Count only the breaker, and 9 in 10 faults look temporary.
As a breaker count sees it · 8,619 breaker operations
What happened to the load · 46,726 reclose attempts
We report a range because the substation sees all the current on a feeder, not every device on it. The low end counts only losses the recording shows. The high end adds load that went missing but that the interval data cannot confirm stayed off.
46,726 reclose attempts, by how much of the loss is confirmed
High confidence 36% · upper bound 53%
Of the reclose attempts that ended with nothing lost, 86% held on the first reclose. The second added 9%; the third, 3%.
Only 14% of successes needed more than one reclose.
19,116 reclose successes with nothing lost, by the reclose that held
Faults restrike, grow and hand off between devices. 9% of fault sequences change configuration mid-fault. When a single-phase-to-ground fault changes, it picks up a second phase 78% of the time.
Three recordings from the fleet:
Tap the chart to enlarge it.
It starts on one phase, spreads to two, then to all three. A downstream recloser keeps closing back in, then locks out, and the load behind it stays off. What would you call it?
Tap the chart to enlarge it.
A fuse clears the fault in half a cycle. The breaker trips anyway, recloses and holds. A breaker count logs a success; the customers behind the fuse lost power.
A fault with a half-cycle peak of 2,717 amps (1,921 A rms, the figure in the table) cleared in half a cycle - faster than any recloser can open. Load current continued on the feeder for an additional 2.5 cycles before the substation breaker operated: the breaker saw the fault current, initiated its tripping sequence, and successfully reclosed for a fault which had already been cleared by a fuse. This is not a temporary fault: the reclose held because the fuse had already removed it, and the load behind the fuse was lost.
Tap the chart to enlarge it.
An arc between two phases climbs, stretches and restrikes. No protective device operates. One label cannot describe it.
An A-B fault initiates through an arc in air. As the arc rises due to heat, it lengthens, increasing its resistance (and decreasing the current). This creates a race condition between overcurrent protection and the ability of the arc to sustain itself in air. If the voltage is unable to sustain the arc, it self-extinguishes, and may (or may not) restrike depending on whether the fault condition is still present.
In storm conditions, faults arrive at 22 times the fair-weather rate. But fair weather fills 81% of the hours, so it still produces 52% of all fault sequences.
Plan for storms. Expect faults every day.
Fault sequences per 1,000 feeder-hours in each kind of weather
Share of hours and of faults: Fair 81.5% / 51.9%; Rain 13.6% / 34.5%; Windy 3.3% / 3.7%; Windy + rain 1.3% / 5.0%; Storm 0.3% / 4.8%. Weather is the hourly reanalysis at the substation, not at the fault.
EPRI's EL-3085 study is still the reference most protection engineers know. Here is what it measured, beside what DFA measures today.
| Measure | EL-3085, 1983 | DFA today |
|---|---|---|
| Sample | about 120 recorded fault events; 50 feeders, 13 utilities, about 18 months | 64,382 fault sequences; 676 feeders, 22 utilities, 2,244 monitor-years |
| Smallest faults seen | very few records below 800 A (trigger levels set high to avoid load) | 50% of sequences peak below 800 A |
| Fault current | line-to-ground average 2,095 A; largest 6,100 A | average 1,080 A, median 800 A (the largest in each sequence); 1,720 A average above 800 A |
| Fault current duration | line-to-ground average 8.35 cycles per event, every opening counted; longest event 73 cycles | median 3.5 cycles (every shot in a sequence); average 19.7, pulled up by a long tail: 10% of sequences run past 49 cycles, 6% past 73 (the 1983 study’s longest) |
| Single phase to ground | 79% of the faults utilities reported | 83% of sequences by the configuration held longest (as EL-3085 typed its events); 79% at the widest configuration each reached |
| One fault per event | about 80% of recorded events (126) | 77% of sequences; 1.47 faults per sequence on average |
| One breaker opening was enough | about 85% of events where the breaker operated (66) | 78% of 8,619 reclose attempts where the substation breaker opened |
| A fuse cleared the permanent fault | 88% of 237 faults utilities reported (they had estimated 59%) | 55% to 64% of sequences that ended with load lost (likely fuse blow; with "fuse unconfirmed") |
DFA connects to the CTs and PTs already in your substation and turns every fault into an answer: what happened, which device acted, and who lost power.
