Explainer

Digital substation architecture in ten minutes

2026-10-29 · 7 min read

A digital substation replaces copper signal wiring with a network, and that single change ripples through everything: what a relay receives, who is responsible for time, how a test is isolated, and what "the wiring is wrong" even means. Here is the architecture in the terms a software engineer needs.

Two buses, two jobs

• The process bus carries sampled values from merging units to protection and control devices, and GOOSE for commands to switchgear. Its requirements are timing and determinism.
• The station bus carries MMS: supervision, configuration, reporting to gateways and control centres. Its requirements are compatibility and scale.
• They are usually separate networks, and mixing them is a design decision with consequences for the worst-case latency of both.

The merging unit changes the question

With conventional transformers, every device has its own analogue input and its own sampling instants. With merging units, one device samples and many consume, so the questions become: what is the time base, what happens when a stream is late, and who detects a missing sample.

This is why synchronisation moves from a nice-to-have to an architectural requirement, and why a device that only synchronises at startup is a liability in a digital station.

What stays the same

Protection philosophy. Interlocking logic. The need for sequence-of-events evidence. The requirement that a test can be performed without endangering the primary plant. None of these are protocol questions, and all of them are the reasons the protocol choices matter.

Where software teams get surprised

• Testing is now a network activity: simulating a merging unit is how you inject a fault, and the tooling is different from a test set with leads.
• Time is a device: a failed grandmaster is an outage of a different kind, and it must alarm.
• Captures matter more: with copper, a wrong wire is visible; with a network, only the bytes tell you what happened.
• Isolation: a device that publishes into the process bus during maintenance can affect others. Design the test mode before you need it.

Questions that come up

Is a digital substation cheaper?

It removes copper and can reduce installation time, and it adds network equipment, engineering effort and a dependence on time synchronisation. The case is usually made on lifecycle and flexibility, not on capital cost.

Can a conventional station be retrofitted?

Partially. GOOSE for interlocking is a common first step because it does not require replacing CT/VT wiring; the process bus is a larger commitment.

What breaks first in practice?

Time synchronisation and multicast network configuration, in our experience — both of which are infrastructure problems that surface as protocol symptoms.

Keep reading