⚡ TL;DR – EV charger load management architecture
Your pilot site works – the chargers come online, the back office talks to them, and the energy bill makes sense. Then you try to repeat that success across 10, 50 or 100 sites, and the cracks appear.
Most large-scale failures are architectural. Sites depend too heavily on cloud-only control, underestimate network and metering, treat OCPP casually, or lock into one vendor. The pilot looks clever mainly because it’s small.
To scale without wasting capacity or painting yourself into a corner, keep safety-critical control local and use the cloud for visibility, billing and fleet logic.
A site-based controller such as LinkRay, coordinated by SiteManager and backed by robust OCPP software, gives you a resilient local/cloud split that can support 100 to 1,000 chargers across an estate.
Why do pilot sites mislead engineering teams?
Pilots hide a lot of sins.
With 6 to 10 chargers on one site, an improvised architecture mostly gets away with it. There’s spare capacity on the its network, one back office to worry about, metering that’s “good enough” because nobody’s testing its limits.
Scale that to 100 chargers and the picture changes.
Every charge point is a networked device talking to a local controller and a back office, while building and plant loads become genuinely variable and substation constraints start to matter. A design that looked tidy at 10 chargers starts throwing intermittent faults, unexplained deratings, and support tickets nobody wants to own.
Pilots survive weak architecture; estates don’t.
What needs to stay local, and what can safely sit in the cloud?
If you take one thing from this article, take this:
the control loop that shares out power in real time and protects your site belongs on the site, not in a data centre three hops away.
When a charger asks for power, it needs to talk to a local controller that can see what the rest of the site is doing right now, with a direct, low-latency link to the chargers and the site meter, so it can enforce hard limits even when the wide-area network is having a bad day. If that decision has to go out to the cloud and back before the site can act, you’ve added risk you didn’t need.
The cloud still earns its place, as the natural home for:
- Estate-level visibility and dashboards
- Billing workflows and payment reconciliation
- Remote firmware updates and support
- Fleet-level optimisation policies over weeks or months
If that connection drops for a few hours, it should be inconvenient, not dangerous. You might delay some reporting or transaction syncing, but you shouldn’t lose the ability to share site power safely between chargers that are still physically present and still drawing load.
LinkRay makes this split explicit. It sits on-site as a local controller between the chargers and the CSMS, enforcing hard site limits and dynamic load balancing while buffering transaction data if the back office goes offline.
When the connection returns, it syncs sessions to the CSMS. There’s no cloud-side decision logic to “replay,” because that logic never left the site in the first place.
Related reading: What happens when cloud-side security fails
Where do load management projects usually go wrong?
Four patterns account for most of the failures we see once a project moves past its first site:
1. Underspecified network and IT infrastructure
Each charger looks like “just another device on the network” on paper. In practice you have tens or hundreds of endpoints that need to talk reliably to a local controller and a back office.
If the switches, routers, VLAN plan or cabling weren’t sized for that endpoint count from the start, you’ll hit bottlenecks and intermittent faults once you scale.
Relying on Wi-Fi for this kind of control is a common symptom. Wireless is convenient, but coverage depends on building materials, distance and interference, and a site that commissions cleanly can still develop dead zones as the environment changes. For hardware that has to work reliably, a wired Cat5e or Cat6 connection remains the safer engineering choice.
2. Weak or missing metering
Dynamic load management depends on seeing what the rest of the site is doing. Without that visibility, you’re managing uncertainty rather than power.
Take a site with 200 amps to share between chargers and other building loads, where a piece of machinery can draw 60 amps at times but there’s no meter at the right distribution point.
The safe response is to reserve those 60 amps permanently, so the charging system behaves as though the site only has 140 amps available, even when the building is barely drawing anything. That’s wasted capacity, not because the supply is too small but because the architecture can’t prove what the rest of the site is doing at any given moment.
Get the metering right and the controller can use the full envelope dynamically instead.
3. Wrong meters, wrong comms
Even where meters are fitted, they’re not always chosen with control in mind.
Some projects pick meters that don’t support Modbus, or don’t support Modbus over TCP/IP, and end up facing expensive rewiring or a weaker control scheme, with a controller that’s blind to parts of the site it’s supposed to manage.
4. Treating OCPP as optional
Designing EV infrastructure today without OCPP is asking for trouble.
Standards aren’t magic, but they’re one of the few practical tools for keeping your options open on charger brands, back offices and future integration work.
Choose a design that only works with one charger brand or one proprietary back office and you might move fast through phase one, but every later change gets harder.
If that vendor discontinues a model, changes firmware support, or exits the market, you’re left with an estate that can’t evolve cleanly.
What does brand-neutral architecture actually buy you?
Real estates rarely offer identical conditions site to site. You inherit different meter brands, mixed AC and DC chargers, and infrastructure specified by a previous team years earlier. Customers merge, sites get divested, back offices change. The odds of finding matching hardware across dozens of locations are close to zero.
A brand-neutral architecture doesn’t promise everything is interchangeable for free. It promises that change is survivable. Support only one meter family and you’ve ruled yourself out of dynamic load balancing on any site that doesn’t use it. Build around one charger vendor and you’re betting your estate on that company’s roadmap and financial health.
An architecture built around OCPP and a brand-neutral local controller gives you time and choice: mixed fleets are easier to support, charger replacements can be staged rather than ripped out, and new meter types don’t force a redesign of the whole stack.
LinkRay is charger-agnostic at the OCPP layer and works with a range of meter brands when properly connected. SiteManager extends that same approach across multiple LinkRays and larger estates, coordinating up to 64 units.
What does a scale-ready architecture look like in practice?
A handful of things tend to be true of designs that scale cleanly. Use this as a self-check against your own plan:
- Network sized for where you're going, not where you are. Switches, cabling and VLAN planning account for the charger count you expect at full build-out, not just phase one.
- Meter placement decided before the trenches close. The team has worked out which distribution points need metering, and specified meters that support the protocols the controller expects.
- Local control and back office on separate paths. There's a clear local path for real-time load decisions and a separate cloud path for visibility and administration, so a WAN outage doesn't stop the site enforcing its limits.
- Chargers and CSMS both speak OCPP, keeping the door open on brand and back-office choice.
- Open enough to tolerate mixed hardware over time. A new charger, meter brand or CSMS can be introduced without replacing the whole control stack.
Within Versinetic’s portfolio, that pattern is per-site LinkRay controllers handling local load and safety, with SiteManager coordinating multiple sites where the estate needs it.
Can you really design for 1,000 chargers?
You can design with 1,000 chargers in mind, but you can’t design your way around site power. Better load management lets you use what you have more efficiently and safely, it doesn’t create capacity that isn’t there.
At high charger counts, incoming supply, site distribution and substation constraints become the dominant issue, well before anyone’s discussing software.
Once the power envelope is tight, every weak decision costs more: poor meter visibility wastes capacity, cloud-dependent control adds latency and risk you don’t need.
A realistic architecture accepts those constraints, uses the available power well, keeps control resilient, and stages growth sensibly. Thousands of chargers is a power engineering problem before it’s a software one.
Are you designing for today's standards or the next five years?
The EV charging market hasn’t settled. OCPP keeps moving forward, grid interaction expectations keep shifting, energy systems get more mixed, and ISO 15118-20 brings new requirements into scope, mandatory for newly installed or renovated public and private chargers in the EU from 1 January 2027 under Regulation (EU) 2025/656.
Design only for today’s minimum compliance and your architecture risks feeling dated well before the hardware reaches end of life. Better to ask what needs to stay stable for the next five years: the local control boundary, the site metering strategy, the communications approach, and the interoperability model.
Get those right and you can absorb the next wave of standards and commercial models without a rebuild. Versinetic’s OCPP software stack and hardware roadmap are being developed with this horizon in mind.
Where this leaves you
Pilots hide architectural weaknesses that estates expose. Keep hard, safety-related control local, use the cloud for billing, reporting and fleet logic, meter properly, choose comms your controller can actually use, and stay brand-neutral.
None of that creates power that isn’t there, but it stops you wasting what you’ve got, and it means your first hundred chargers don’t become the reason your next nine hundred are a rebuild.
If you’re reviewing a rollout plan, a depot design, or a charger product roadmap that now needs to scale, put the architecture under pressure before the estate gets bigger.
A short engineering review of your local/cloud split, metering and comms strategy, and interoperability assumptions can save a much nastier redesign once the weaknesses show up in the field.
Get a second opinion on your architecture
FAQs
How a charging site measures available power, shares it between chargers, and decides which control functions stay local versus which sit in the back office, covering chargers, local controllers, site meters, communications, OCPP and the CSMS boundary.
Because the local control path keeps working when internet or cellular links fail, so the site can go on managing charger power safely even if the back office connection drops for a period.
Static load management works to fixed limits per charger or group, regardless of what the rest of the site is doing. Dynamic load management adjusts charger output in real time based on live site load and remaining capacity, and that only works well if the architecture can see the right metering points.
In most serious installations, yes. Without proper metering, the controller usually has to reserve capacity for worst-case building load, so chargers end up with less usable power than the site may actually have available.
They can in some cases, but for hardware that has to work reliably, a wired Ethernet connection is usually the safer engineering choice. It’s more predictable, easier to fault-find, and less exposed to coverage issues than Wi-Fi on its own.
Ritchie Perry is an engineer working across Versinetic and its sister company, embedded systems engineering consultancy ByteSnap Design.
With more than 30 years of experience in the electronics industry, he began his career as a sound engineer before moving into professional audio electronics, working with leading brands including Midas, Klark Teknik, Behringer and Lab Gruppen.
Recognising his potential, an early employer supported Ritchie’s professional development by funding a BEng in Electronic Engineering at Birmingham City University, where he graduated with first-class honours.
Ritchie divides his time between in-house production testing, electronics engineering, and on-site work with clients to deliver practical engineering solutions. Outside of work, he is a keen guitarist and performs with several tribute bands.
Further reading
- Introduction to Load Management – Cenex’s independent guide to EV load management levels and principles
- Open Charge Point Protocol – Open Charge Alliance, the standards body that develops and maintains OCPP
- Regulations: electric vehicle smart charge points – GOV.UK guidance on the Electric Vehicles (Smart Charge Points) Regulations 2021, including the requirement that a charge point must keep charging even if it loses its network connection
- Commission Delegated Regulation (EU) 2025/656 – the EUR-Lex text confirming ISO 15118-20 becomes mandatory for newly installed or renovated public and private Mode 3 charging points from 1 January 2027
- Understanding Load Management Systems for EV Charging Infrastructure – Facilities Management Insights on static vs dynamic load management from a building-owner perspective
- What is dynamic load balancing for EV chargers? – control principle and benefits
- LinkRay: flexible subscription-free local load manager – site-based controller details
- SiteManager: advanced EV charger load balancing system – coordinating multiple LinkRays at scale
- EV smart charging software, OCPP & software licensing – protocol stack and CSMS integration
- UK EV Charging Standards 2026 Guide
