Pad-Mount Fuse, Relay TCC & Arc-Flash

Every inverter skid in a solar or BESS plant feeds a pad-mounted step-up transformer: typically about 3–7 MVA, with an LV winding of about 480–800 V on the inverter side and 34.5 kV on the collector side (representative ranges). A large plant might have a hundred or more of them, loop-fed by underground cable, about 6–12 to a feeder.

The protection problem looks simple: if something inside the skid fails, disconnect it. But think about which way fault current flows. The inverter itself contributes little: typically about 1.0–1.5 pu of rating, set by the OEM current limit, after a sub-cycle transient that can be higher. The real energy comes backwards from the 34.5 kV collector, which is tied to the whole plant and the transmission grid behind it.

Where the fault is decides who clears it. For a fault inside the inverter, downstream of its LV AC breaker, that grid-fed current flows through the LV breaker, which is rated for the LV fault duty and clears it selectively. The difficult zone is the short stretch between the transformer LV terminals and the LV breaker: the LV bus, cable and terminal compartment. Nothing on the LV side can interrupt a fault there, so it must be cleared from the 34.5 kV side. That is the fault this module is about.

Traditionally that device is a fuse pair: an oil-immersed weak-link expulsion fuse in series with backup current-limiting fuses. Fuse-only protection, sometimes with a skid-mounted switch or vacuum fault interrupter (VFI), is still the most common baseline. Fuses are cheap and reliable, but they are slow at moderate currents, so a 690 V arc on the LV bus can burn for seconds. One option this module explores adds a microprocessor relay on the 34.5 kV bushing CTs that can trip the upstream feeder breaker by fibre direct transfer trip (DTT), relegating the fuses to backup. Its cost is selectivity: the feeder breaker carries every skid on the feeder (often several tens of MW), so §04 shows how to use it only when it is needed.

A TCC plots operating time against current on log–log axes. Each device is a curve: anything to the right of and above it operates. Coordination means arranging curves so the device nearest the fault always operates first, with margin.

• Inverse-time relays (51) follow the IEEE C37.112 equation. Pickup sets where the curve starts and the time dial (TD) slides it vertically.
• Instantaneous / definite-time (50) stages are horizontal lines: above pickup, trip after a fixed delay.
• Fuses have two curves: minimum-melt (element starts to melt) and total-clear (arc extinguished).
• Damage curves (IEEE C57.109) show how long the transformer can carry a through-fault before thermal or mechanical damage.

The weak-link expulsion fuse (often a dual-sensing bay-o-net link in oil) handles overloads and low-to-moderate internal faults. It cannot safely interrupt very high currents. The partial-range (backup) current-limiting fuse handles those. In this design it is two 200 A CLFs in parallel. A CLF forces the current to zero within the first half-cycle and limits the let-through $I^2t$ that could rupture the tank.

How the pair is sized. The bay-o-net link is chosen from the fuse vendor's table for the transformer kVA and voltage. It must carry the allowed overload and cold-load current, ride through energization inrush (about 12× FLA for 0.1 s), and melt fast enough to protect the transformer. Here a 100 A dual-sensing link (representative) melts above about 180 A, roughly 2.7× FLA. The backup CLF is then chosen from the vendor's matched-pair table, so that its minimum interrupting current lies below the crossover with the link and below the link's own maximum interrupting rating. That requirement, not the transformer load, is why the CLF rating looks large next to a 67 A unit. Use one CLF where the vendor table allows it.

The pair must be matched:

• Below the CLF's minimum interrupting current a backup CLF must never be asked to clear, so the weak-link has to operate first there.
• The crossover, where the weak-link total-clear curve meets the CLF min-melt curve, must lie above the CLF minimum interrupting current. In the lab the cyan band hands over to the violet curve at about 2.4 kA, above the 1.4 kA minimum break. Above the crossover the CLF also protects the weak-link from currents beyond its own interrupting rating.

A multifunction relay on 34.5 kV bushing CTs at each pad-mount provides an inverse-time 51 set below the weak-link minimum-melt curve, plus two 50 stages placed well below the fuses at arcing-fault current levels. For a fault on the LV bus it sends a DTT over fibre to open the 34.5 kV feeder breaker at the substation. Total clearing is roughly one cycle of relay time, a few milliseconds of channel time, and a 3–5 cycle breaker: about 0.1 s instead of seconds. A selective scheme adds the wait for the LV breaker (below), so about 0.15–0.2 s, unless an LV arc-flash relay identifies the arc.

The starting settings look reasonable, but closing the feeder breaker trips this pad-mount. Use the lab to reach a fully coordinated design for an LV-bus fault: the relay is primary for everything up to the 690 V bolted fault, the fuses are backup, the transformer is protected (including against a bolted LV ground fault, checked against the 0.58-shifted damage curve), and inrush is secure.

When the 34.5 kV feeder breaker closes, every pad-mount on the loop magnetizes at the same instant. Each core can be driven into saturation by residual flux and the point-on-wave of closing, drawing a peaky, one-sided current: conventionally modelled as ~25× FLA for 0.01 s and ~12× FLA for 0.1 s. With the arcing-fault 50 stage set at a few hundred amps, inrush looks like a fault. In practice only the 12× point binds: even a zero-delay 50 stage needs about one cycle (16 ms) to operate, so it cannot trip within the 0.01 s of the 25× point.

Inrush, however, is rich in even harmonics, especially the 2nd. A fault on a 690 V bus is nearly sinusoidal. The relay measures the 2nd-harmonic ratio $I_{h2}/I_{h1}$ (not to be confused with the negative-sequence $I_2$ used elsewhere in this course) and inhibits (torque-controls) the overcurrent elements while it exceeds a threshold, typically 15–18%. Fibre status from the upstream breaker can further arm the block only during the energization window.

On overhead distribution, most faults are temporary (lightning, a branch), so reclosing restores service. A collector feeder is underground cable, and a cable or splice fault is a permanent insulation failure. Reclosing into it only adds damage and another fault-current pulse to the cable shield and the pad-mounts.

The inverters add a second reason. During the dead time the downstream skids may keep the islanded feeder energized briefly, or ride through in an indeterminate state. Reclosing can then connect two unsynchronized sources out of phase, producing a transient that can damage transformers and power electronics. With delta-HV pad-mounts the island is also ungrounded: the substation grounding bank is left on the bus side of the open breaker, so a ground fault on the feeder drives the sound-phase voltage toward √3 until the inverters stop (see the feeder-island callouts in Modules 1.2 and 2.2). Collector feeder relays therefore trip and lock out until the cable has been located, tested and repaired.