Making the Substation Intelligent
The substation at the end of your road has probably been there since before you were born. The transformer inside has been replaced once, maybe twice. The protection relays many more times. The fence, the foundations, and the right of way are the only things that have actually been there since the 1960s.
That sentence could be written about a substation in Warwickshire, in Bavaria, in Jiangsu province, or in Ohio. The technology inside varies; the pattern does not. Substations everywhere outlive the equipment bolted into them, and the equipment outlives several generations of the digital systems that protect it. Modernising them is a retrofit problem, not a rebuild problem — and the international standards that define how to do it (IEC 61850 for the digital substation, IEC 62443 for the security architecture, IEC 62351 for protocol-level protection) are the same everywhere.
The scale of what’s happening is easy to miss. China’s State Grid Corporation has mandated IEC 61850 for all new substations since around 2009 and has deployed thousands of digital substations with full process-bus architectures. RTE, the French transmission operator, has been running virtualised protection in production since late 2023 on an open-source hypervisor. In the US, FERC has just cleared the regulatory path for virtualised substation infrastructure.
I come at this from a cloud-infrastructure background. For most of my career, the relevant timescales have been the release cycle and the latency budget. In the grid, the relevant timescales run from milliseconds at one end to decades at the other — and the interesting work happens in the bridge between them. I live in the UK, so the grid I walk past and the public documents I started reading are British — Ofgem price controls, National Grid business plans, NCSC cyber frameworks. What follows uses the UK as a worked example, but the IEC standards that underpin everything are global, and the engineering trade-offs are the same whether you’re retrofitting a 275 kV substation in England or a 500 kV substation in Anhui.
The ratio between the longest-lived and shortest-lived components on a single site is roughly four to one. Civils last 60–80 years, primary plant 30–40, protection 15–20, SCADA faster still. You don’t wait for a transformer to fail before upgrading the protection that sits on top of it. You retrofit the secondary systems — relays, merging units, networks, increasingly the compute — on a software cadence while the primary plant carries on through its own much longer replacement cycle.
A protection relay has three milliseconds to decide whether to open a 400 kV breaker.
Not three seconds. Not thirty milliseconds. Three. That’s the figure in IEC 61850-5 for a Type 1A trip message — the signal that says “a fault has been detected, isolate this piece of grid right now.” Miss it and the fault propagates. Components melt. The lights go out for a postcode. In a bad case, a cascade takes out a region — as the Iberian blackout of 28 April 2025 demonstrated when protection trips cascaded faster than frequency response could compensate, cutting power to sixty million people across Spain and Portugal within seconds.
Once you’ve got those frames — sites that outlive equipment, equipment that outlives software, and software with a millisecond budget — everything else about digital substations starts to make sense.
The bit nobody outside power engineering talks about
A substation is where voltage steps up or down and one set of circuits hands power to another. Walk around one and you’ll see bays — each bay a self-contained unit with a circuit breaker, disconnectors, instrument transformers, and a line or transformer hanging off it.
For decades, each bay had its own rack of protection equipment: individual relays doing individual jobs, wired with copper trunks back to a central control room. Fixed-function devices. Thousands of them per site. Expensive to install, expensive to upgrade, and for the most part, deaf to each other except through laboriously-configured hard wiring.
The industry response since the early 2000s has been IEC 61850 — a standard that does for the substation what Ethernet did for the office: replace bespoke point-to-point copper with a shared network, and let the devices publish and subscribe to each other’s data.
That’s the hinge on which everything else turns.
Transmission versus distribution
Not every substation is the same beast. The grid has two broad tiers and they look quite different when you walk around them.
Transmission substations sit at the highest voltages — 275 kV and 400 kV in Great Britain, typically 230 kV and above elsewhere. They are fewer in number, physically enormous, and connect generators, interconnectors, and major load centres into a meshed network. A single transmission substation can have forty or more bays and feed multiple cities. Loss of a transmission bus is a system event: the protection schemes have to clear faults in tens of milliseconds, and the network above them is engineered to survive any single component failure (the “N-1” rule). Operationally, transmission is the domain of the transmission system operator (TSO) and its Energy Management System (EMS) — the software that runs state estimation, contingency analysis, and the real-time security assessment of the transmission grid.
Distribution substations sit below, stepping voltage down through a hierarchy: primary distribution at 132 kV or 33 kV, secondary at 11 kV, and finally the ubiquitous low-voltage transformers in roadside cabinets and pole-mounted units that feed streets and houses. There are hundreds of thousands of them in a typical country, most unattended. The network below them is largely radial. Failure of a single distribution substation can darken a neighbourhood, but it is not a grid-stability event. Operationally, distribution is the domain of the distribution system operator (DSO) and its Advanced Distribution Management System (ADMS) — increasingly important now that rooftop solar, batteries, and EV charging push bidirectional power flows down through the distribution hierarchy in ways the original 1960s design never anticipated.
The data substrate that lets ADMS and EMS exchange network models with each other and with the wider grid is the Common Information Model (CIM) defined by IEC 61970/61968. The data substrate for the DER endpoints themselves is IEC 61850-7-420, the part of 61850 that extends the logical-node catalogue to solar PV, batteries, EV chargers and the rest.
IEC 61850 applies to both tiers, but the rate of modernisation is uneven. Transmission substations get rebuilt carefully, often in greenfield extensions rather than rip-and-replace. Distribution substations — where automation has historically been patchy — are where a lot of the current investment is actually landing, driven by the DER integration problem and by regulatory pressure to observe and control what is happening out at the edges of the grid.
Refresh and retrofit
A substation is not a single asset. It’s a stack of components with very different design lives — and very different replacement cadences — bolted to the same site. The site itself is typically the only thing that survives intact from one decade to the next; everything inside refreshes on its own clock.
| Component | Design basis | Typical UK in-service age | Refresh driver |
|---|---|---|---|
| Civil works (foundations, control buildings) | 60-80 yr | Often original (50-60 yr) | Site reconstruction only |
| Power transformers | ~40 yr (IEEE C57.91) | 25-40 yr; refreshed via RIIO | Condition score (NARM) |
| Switchgear (AIS / GIS) | 30-40 yr | 25-40 yr | Replacement programme |
| Electromechanical relays | 30-50 yr | Some 1960s-70s units still in service | Project-driven retrofit |
| Numerical / digital relays | 15-20 yr | 2-3 generations per transformer life | Vendor firmware end-of-life |
| SCADA / RTU hardware | 10-15 yr | Effectively shorter | OS / protocol obsolescence |
In the UK, this lifecycle mismatch is managed through Ofgem’s RIIO price-control cycles. RIIO-T1 (2013–21) refurbished 74 transmission substations — about 22% of the UK transmission estate of roughly 330 sites — most of that work being secondary-system replacement on existing primary plant rather than greenfield rebuild. RIIO-T2 (2021–26) was bigger. RIIO-T3 (2026–31) is bigger still at £35bn. Other countries have different funding mechanisms — China’s five-year state plans, US rate cases before FERC and state PUCs, European regulatory periods — but the pattern is the same everywhere: a continuous flow of partial refreshes rather than the rare grand rebuild, and the unit of replacement is one or two asset classes at a time, not the whole site.
That’s the problem. Sites that outlive equipment. Equipment that outlives software. Software with a three-millisecond deadline. And a continuous flow of partial refreshes — one or two asset classes at a time, never the whole site — stretching across decades.
The next post picks up where the table leaves off: the stand-alone merging unit that bridges a 1975 transformer to a 2025 protection scheme, the Ethernet network that carries the signal, and the engineering question of whether you still need one relay per bay once everything is digital.
References
Standards
Asset lifetimes and refurbishment
- Ofgem — RIIO-T1 closeout decisions for transmission network operators (2022)
- IEEE C57.91 — Guide for loading mineral-oil-immersed transformers and step-voltage regulators