Surges & arresters · Backflashover Lab

When the shield wire is struck: backflashover

The shield wires did their job and caught the stroke. The current now has to get to earth, through the tower and its footing. On the way it lifts the whole tower to millions of volts. The phase conductors are lifted much less, so the insulators see the difference. If it lasts long enough, a leader bridges the horn gap from the tower side: the flashover runs "backwards", from the earthed tower to the live conductor.

  1. 1Stroke on the towerThe current splits: some goes down the tower, the rest runs out along both shield wires.
  2. 2Tower risesThe tower top jumps to I × (ZT ∥ Zg/2 ∥ Zch), several MV for a big stroke.
  3. 3Footing answersAfter 2H/v the wave reflected by a low footing resistance comes back up and pulls the top down.
  4. 4Neighbours answerAfter 2L/c the adjacent towers send relief waves back along the shield wires.
  5. 5Phases follow, partlyCoupling lifts each phase by only 30–45% of the tower. Arresters that conduct tie their phase to the tower.
  6. 6Leader across the hornsThe insulator sees tower − phase − power-frequency voltage. A leader grows; if it bridges 1.44 m, the phase flashes over.
t = 0.00 µs

Tower, shield wires and phase S

spans drawn shortened
shield-wire voltagephase S voltageleader / flashoverarrester conducting

At the struck tower

both circuits are mirror images
PhaseArresterPeak |u|LeaderFlashoverGapIarrQ

Tower top, footing and phase S

kV
tower topfootingphase S conductor

Voltage across the insulators

|arm − phase|, kV
RSTgap sparkover912 kV calibration

Leader across the horn gap

fraction of 1.44 m

Arrester and footing current

kA
RST arrester (where fitted)footing ÷ 10
Critical current

How big must the stroke be to flash the tower?

Each point is a search over the stroke current for the smallest one that flashes any insulator at this tower, with the angle, front and gap above. Arrester schemes are compared at the same settings.

no arrestersR/T (as installed)R/S/Tthis tower's R₀

At the selected footing resistance

SchemeIcFirst to flashStrokes above Ic
computing…
What to take away

Six lessons

Why "back"?

In a direct strike the conductor rises first and flashes to the tower. In a backflashover the tower rises first and flashes to the conductor. The same insulator fails, but from the other end, and the cure is different: lower footing resistance or arresters, not better shielding.

The footing sets the level, the span sets the time

The tower top peaks before the footing reflection returns (2H/v ≈ 0.4 µs). After that, the voltage is roughly the current times the footing resistance in parallel with the shield wires. At T20 (θ = 120°), lowering R₀ from 66 Ω to the June 2026 value of 15.75 Ω raises the critical current from about 41 to 59 kA without arresters, and from 85 to 111 kA with R/T arresters. Soil ionisation already does part of the job: switch it off and 66 Ω gives only 31 kA.

Arresters move the risk to phase S

Without arresters the phase with the worst mix of coupling and power-frequency voltage flashes first. With R/T arresters those two phases follow the tower, and their coupling helps phase S. The critical current roughly doubles, but what remains falls on S. In the paper, −96 kA flashed C2-S at T20, and at −120 kA 40 of 48 flashed paths were on phase S.

A peak voltage is not a flashover

Choose "Paper H1": phase S sees about 0.93 MV, above the 912 kV calibration level, yet the leader hardly starts before the voltage falls. The paper found the same at 1.04 MV: the leader grew 0.062 m of 1.44 m. A fixed-CFO check would call this a flashover.

The phase angle matters

A negative stroke drives the tower negative. A phase near its positive peak sees the largest difference. At θ = 120° phase S is at +122 kV, the angle the paper uses for its stress cases. Without arresters, R flashes first at θ = 0°, S at 60–120° and T at 180–300°. With R/T arresters it is always S.

It did not explain the 2 May 2026 outage

The recorded −68.6 kA stroke does not flash C1-S here or in the paper, and arresters would have counted. That is why the paper looks for a direct strike on the middle phase instead. See the Relief Wave Lab and the 3D Strike Lab.

Model behind the animation

Lines

Eight conductors (two shield wires at ±3.2 m; R, S, T of both circuits at ±4.5 m, 2.9, 7.6 and 12.3 m below the shield wires). The two circuits are mirror images, so the surge-impedance matrix Zij = 60 ln(Dij/dij) folds to 4 × 4. Multiconductor Bergeron line, lossless, c = 300 m/µs, Δt = 5 ns. Back and ahead spans from the thesis (T20: 302.2 m and 645.6 m). At the adjacent towers the shield wires go to earth through their footing resistance and every conductor continues, matched.

Tower and footing

Single tower surge impedance 178.9 Ω (paper, T20), velocity 0.8c. Footing R(i) = R₀ / √(1 + i/Ig), Ig = E₀ρ / (2πR₀²), E₀ = 400 kV/m. The cross-arm voltage is interpolated between the footing and the top.

Stroke

Heidler current (n = 5) with front time T₁ = 1.67·(t90 − t30) as chosen (1.2 µs for the paper's 1.2/50 µs screening wave) and a 50 µs tail, with a parallel 400 Ω channel, at the tower top.

Insulator and arrester

u = Varm − (Vphase + vpf), vpf = 122.5·cos(θ − 120°k) kV for R, S, T. Leader progression dl/dt = k·u·[u/(g − l) − E₀], g = 1.4395 m, E₀ = 634 kV/m, k = 1.8 m²/(kV²·s); flashover when l = g, then a 10 Ω arc. EGLA: gap closes at |u| ≥ 570 kV, then u = 306 kV·(i/10 kA)^(1/14).

Against the paper's EMTP-RV model (T20, R/T arresters, 1.2/50 µs)

−69 kA: tower top about 2.5 MV (paper 3.1 MV at the attachment point), footing 1.46 MV (paper 1.44 MV); R/T gaps spark over at 0.82–1.02 µs (paper 0.62–1.02 µs); no flashover at θ = 0–270° (paper: none). At θ = 90° phase S reaches 0.93 MV without flashover (paper 1.04 MV, leader 0.062 m). −96 kA, θ = 120°: phase S flashes, at 2.6 µs and 0.86 MV (paper: C2-S at T20, 1.70 µs and 1.29 MV). Critical current with R/T arresters about 85 kA, inside the paper's bracket of 69–96 kA. The tower top is lower and the S flashover later than in the paper, because this model has a single-section tower without cross-arm inductance, the stroke at the tower rather than on one shield wire, and no ground-return loss.

Simplifications: stroke at the tower, not mid-span; both circuits identical, so C1 and C2 flash together; adjacent towers are resistances; no corona; no arresters or insulators modelled at the adjacent towers. Treat it as a teaching model.