Most site power decisions start with a number someone remembers from another project: this location got a 120, that one got a 60, so this one probably needs something in between. It is a fast way to reach a figure and a slow way to reach a right one. The power a site actually needs is produced by four quantities — energy per session, sessions per bay, how many vehicles overlap, and how much of the nameplate the unit really delivers — and all four are measurable before anything is ordered.
What follows is the calculation in the order it should be done, with a worksheet at the end that you can fill in with your own site data.
Nameplate power is the last number you need
Installed power is an output of the sizing exercise, not an input to it. Starting from a nameplate figure quietly fixes two other decisions before you have looked at the site: how long a vehicle stays connected and how many bays you build. Both of those belong to the operator or the site, not to the charger.
The useful starting point is a service promise: how long should a driver expect to wait, and how long is the vehicle parked anyway. Everything else follows arithmetically.
The four inputs that decide the answer
| Input | What it means | Where to get it |
|---|---|---|
| Energy per session | The kWh a typical visit actually adds, not the size of the battery | Your own session history, or the vehicle mix you expect to attract |
| Dwell time | How long the vehicle occupies the bay, charging plus overstay | Site observation, parking policy, the reason the driver stopped |
| Arrival rate at peak | Vehicles per hour in the busiest window, not the daily average | Traffic counts, existing sessions, comparable sites nearby |
| Delivered fraction | Average power the unit really delivers as a share of its nameplate | Charge logs, including taper and shared-power behaviour |
Three of these are properties of the site. Only the fourth is a property of the hardware, and it is the one sales conversations talk about first.
Step one: energy per session
A driver does not arrive empty and leave full. They arrive at whatever state of charge their journey left them with, and they leave when they have enough for the next leg — or when the parking meter expires. The energy per session is therefore a behavioural figure:
Es = usable capacity × (SOCend − SOCstart)
Both SOC values come from where the site sits, not from the vehicle brochure. A highway stop and an office car park produce very different numbers from the same car.
If you have no session history, take the median of a comparable site rather than the maximum. Sizing on the largest vehicle that might arrive is how sites end up with power they pay for monthly and use a few times a week.
Step two: how many sessions a bay can serve
A bay's capacity is set by the time it is occupied, and occupancy is charging time plus everything else: queuing, plugging in, walking away, overstaying. The charging portion is the only part power affects.
tcharge = Es ÷ Pavg Nbay = H ÷ (tcharge + tother)
H is the hours the bay is usable in a day; tother is turnover time.
This is where most sizing arguments resolve themselves. If tother is larger than tcharge, doubling power buys little throughput, because the bay is empty for reasons the charger cannot fix. If tcharge dominates, power is the lever — and the site has a queue problem, not a parking problem.
Step three: concurrency is what the transformer sees
Total daily energy tells you the electricity bill. Concurrency tells you the connection. The relationship is the one used in queueing theory and it is short enough to use on site:
C = λ × T
λ is arrivals per hour in the peak window, T is the average time a vehicle occupies a bay. C is the average number of vehicles connected at once during that window.
C is an average, so plan for the peak above it rather than at it. A site designed exactly at C spends its busiest hour queueing, which is precisely the hour that sets the driver's opinion of the network. Add a margin you can defend from your own arrival data, and design the electrical capacity around that figure.
Step four: installed power is not delivered power
Every nameplate number behaves like a ceiling, and several ordinary conditions keep a session well below it: the vehicle's own acceptance limit, the taper above a certain state of charge, the battery being cold, and power sharing when two connectors are in use on one cabinet. Their combined effect is captured in one ratio:
Pinstalled = (C × Pavg) ÷ f
f is the fraction of nameplate the unit delivers on average. Pull it from logs, not from a datasheet.
For a highway site where sessions are short and taper is brief, f is relatively high. For a site where vehicles sit for hours, most of the session happens in the taper region and f falls — which means the same delivered energy needs more nameplate, or a different charging strategy. On long-dwell sites, a lower-power unit used for the whole stay often delivers more useful energy per unit of connection capacity than a high-power unit used briefly.
Three site types, three binding constraints
| Site type | What limits throughput | Sizing emphasis |
|---|---|---|
| Highway service area | Charging time, then bay count. Arrivals come in waves. | Power per bay first, then concurrency at the wave peak. See 120 kW and 180 kW DC units. |
| Retail and destination | Parking turnover, not charging. The car is there anyway. | Energy per bay across the visit; moderate power across more bays beats high power on few. Mid-power units such as the 60 kW DC station fit this pattern. |
| Workplace | The working day. Everyone arrives within the same hour and leaves within the same hour. | Energy over a long window, with load management against the building's existing supply. AC units such as the turbo wallbox are usually the better instrument. |
The retail and destination layout and the workplace layout show how the same electrical capacity is distributed differently once dwell time changes.
The mistake that keeps repeating
Sizing for the busiest ten minutes of the year is the most expensive way to solve a problem that lasts ten minutes. The alternative is not to undersize, but to separate two decisions that get merged: connection capacity and delivered power.
Size the incoming supply around concurrency at a peak you can document, with a margin you can explain.
Size the number of bays around dwell time, because bays are what drivers experience as queueing.
Use load management where the supply is the constraint and the dwell is long, rather than buying a larger connection you need for a short window each day.
One consequence is worth stating plainly: on a long-dwell site, adding power to a bay often raises the connection requirement without raising the number of vehicles served. The bay is occupied either way; the extra kilowatts just arrive faster and then sit idle in the taper. That is a capital decision disguised as a performance upgrade.
The worksheet
Fill these in before requesting a quote. The numbers are yours; the sequence is what makes the result defensible.
Median energy per session, from logs or a comparable site
Median dwell time, split into charging and non-charging time
Peak-hour arrival rate, taken from the busiest window rather than the daily average
Concurrency C, then the peak above it that you are willing to design for
Delivered fraction f, measured rather than quoted
Installed power from the formula, then compared against the supply you already have
If the installed figure lands far above the existing supply, the next question is not which charger to buy but whether to reshape demand — by spreading sessions, capping per-bay output at peak, or adding storage. That decision is cheaper to make on paper than on a switchboard. The technical documentation for any unit lists its input ratings, which is the number the electrical designer needs once the sizing is settled.
Put your own site data through the four steps
If the worksheet produces a figure that sits awkwardly against your existing supply, that is the point worth discussing — before a layout is drawn, not after.
Contact our team