Reconstructing an outage from the evidence
At about 13:49 on 2 May 2026 circuit C1 tripped for a phase-S fault and reclosed. The lightning-location system had just recorded a −68.6 kA stroke near the line. The obvious answer is a backflashover. The burned horns, the silent arrester counters and the models say otherwise. This page follows the paper's reasoning: turn each observation into a test, run every candidate mechanism, and keep only what passes.
What was recorded
Lightning-location strokes near T17–T22
| Time | Peak | Position |
|---|---|---|
| 13:47:39 | −43 kA | north of T19–T20, 0.79 km from T19 |
| 13:48:55 | −68.6 kA | 226 m NE of the line, a third of the span from T20 |
| 13:48:56 | −26.9 kA | 718 m SW of the line, near T21 (211 ms later) |
| 13:50:43 | −18 kA | 0.26 km from T17, outside the span |
| 13:52:13 | −43 kA | 0.77 km from T17, outside the span |
Paper, Table V. The two highlighted strokes coincide with the fault.
What the inspection found, and what each finding demands
| Observation | Constraint on a candidate | |
|---|---|---|
| O1 | Burned C1-S horns at T20 and T21 | C1-S flashes at both towers |
| O2 | No other arcing in T16–T25; C2 unaffected | No other path flashes |
| O3 | R and T counters unchanged at T20, T21 | No registered R/T arrester pulse |
| O4 | Single-phase fault, successful reclose | A transient flashover with power-follow arc |
| O5 | Largest recorded strokes −68.6 and −26.9 kA | A shield-wire stroke is not above the record; a weak stroke may be missing |
Test every candidate
Each row is computed now in your browser with the Backflashover Lab model (shield-wire strokes, 1.2/50 µs) or the Relief Wave Lab model (direct strokes on C1-S, 0.4/14 µs, θ = 300°), and with the span screening for the attachment limit. The paper's EMTP verdict is in the last column.
| Candidate | Model: paths that flash | R/T arresters | O1 | O2 | Current plausible? | Here | Paper |
|---|---|---|---|---|---|---|---|
| running… | |||||||
The backflashover model treats both circuits alike, so a phase-S flashover appears on C1 and C2 together; that alone violates O2. The direct-strike model has a single conductor and no R/T arresters, so O3 there rests on the paper's EMTP result (no R/T conduction up to 12.5 kA).
Where and how strong could the stroke have been?
A direct stroke on C1-S matches only if it flashes both towers (O1) and is weak enough for its leader to reach phase S at that point (I ≤ Imax). Each cell is one simulation. Green cells satisfy both.
Verdict
What the lightning record can and cannot say
- Strokes of tens of kiloamperes are detected with high efficiency, so −68.6 kA bounds the current available for a backflashover.
- Weak strokes are the ones most often missed, peak currents carry errors of about 13%, locations about 0.3 km, and about one flash in three reaches the ground at more than one point. An unrecorded 10 kA stroke is entirely possible.
- Taking the largest nearby stroke as the cause would have pointed to backflashover, and so to better earthing or more R/T arresters. Neither protects phase S.
How the checks are made
The paper's EMTP-RV model has eight frequency-dependent conductors, tower sections and both circuits; these teaching models are simpler, and their flashover times differ by a few tenths of a microsecond. The pattern of passes and fails is what matters.