Protection relays using 5G for inter-substation communication to enhance selectivity

(NSA) and Standalone (SA)—can meet the latency and reliability requirements for inter‑substation protection schemes such as tele-protection.

This study evaluates the feasibility of using 5G for upstream tele‑protection in medium‑voltage radial networks, where a block signal must reach the upstream relay within 100ms to ensure proper selectivity.

GOOSE messages defined in IEC 61850, normally exchanged over Layer‑2 networks, were used to transmit the block signal. Since GOOSE is not routable, VXLAN tunnelling was applied to encapsulate

Layer‑2 frames for transport across 5G Layer‑3 networks. The methodology involved both laboratory tests using a dedicated 5G‑SA indoor cell with local breakout, and field trials across two substations located 12 km apart, using live public 5G‑NSA and 5G‑SA networks. All traffics were captured at switch level, enabling precise end‑to‑end latency measurements.

In the controlled laboratory environment, 5G‑SA performance was stable, with mean latencies around 16ms and a maximum observed latency of 20ms—well below the 100ms requirement. Field tests, however, initially showed large latency variations, especially on 5G‑NSA, with peaks exceeding 100ms.

To address radio‑layer timers such as DRX, background traffic was introduced at 20ms intervals. Additional tuning, including mapping GOOSE traffic to a high‑priority 5Qi in 5G‑SA, significantly improved performance. After adjustments, both networks achieved similar median latency in the 30– 35ms range, but the tail latency differed substantially: 5G‑NSA exhibited peaks up to 387ms, whereas 5G‑SA remained stable below 55ms. From a protection perspective, this tail behaviour is critical, as even 1% of delayed packets represent over 3.5 days per year outside acceptable limits.

Overall, the results show that 5G‑SA meets the protection requirements for this trip‑blocking use‑case, while 5G‑NSA does not, due to high latency dispersion. Future work includes analysing first‑arrival

GOOSE latencies during status changes and evaluating r‑GOOSE, which removes the need for VXLAN and may reduce delays further.