Short answer
Correlated EV charging transformer loading differs sharply from estimates that treat each car independently. If each of 30 homes on a transformer has a 7 kW charger and each is charging at 19:00 with probability 0.2, independently, the expected load is about 42 kW and the 99th-percentile load about 78 kW. But charging is rarely independent. Arrival home after work clusters sessions; time-of-use tariffs or smart-charging defaults start many sessions at the same minute. If a tariff makes 70 % of cars start at 23:00, the load is about 147 kW, beyond a 100 kVA transformer. Managed charging can spread sessions and protect the transformer, but if the management system fails or is overridden, sessions return to synchronised starts. Planning should model arrival patterns, tariffs and control failure as scenarios, not single-car probabilities.
Independent case
| Case | Cars | Load |
|---|---|---|
| Independent, 99th percentile | ≈ 11 | ≈ 78 kW |
| Tariff-synchronised | ≈ 21 | ≈ 147 kW |
| Managed | ≈ 8 | ≈ 56 kW |
Illustrative example. 30 homes; 7 kW chargers; probability of charging at 19:00 is 0.2 each. Values chosen for demonstration.
mean = 30 × 0.2 = 6 cars → 42 kW
σ = sqrt(30 × 0.2 × 0.8) ≈ 2.19 cars
99th percentile ≈ 6 + 2.33 × 2.19 ≈ 11 cars → ≈ 78 kW
Correlated case
With a time-of-use tariff starting at 23:00 and default charger timers:
30 × 0.7 = 21 cars → 147 kW
The binomial estimate no longer applies because sessions are not independent. Even when baseline night load is low, ambient temperature, initial thermal state and available cooling remain relevant to loading capability. Synchronous charging may keep loading above rating for hours.
| Case | Cars charging | Load |
|---|---|---|
| Independent, 19:00 | ≈ 6 mean, ≈ 11 at 99th | ≈ 42 / 78 kW |
| Tariff-synchronised, 23:00 | ≈ 21 | ≈ 147 kW |
| Managed, spread over night | ≈ 8 | ≈ 56 kW |
| Management failure | ≈ 21 or more | ≥ 147 kW |
Sources of correlation
- Arrival time clusters on weekdays.
- Tariffs with sharp price steps.
- Default timer settings in chargers and vehicles.
- Weather affecting heating and range.
- Events such as holidays.
Managed charging
- Local control at the transformer, or utility signals, spread sessions.
- Failure cases: communication loss, override, firmware fault; behaviour on failure (charge immediately or wait) decides the risk.
- Planning should include at least one failure scenario with stated probability or duration.
Transformer assessment
Combine EV scenarios with existing load and heat pumps; see planning transformer capacity with electrification scenarios. Daytime EV charging may absorb PV, interacting with hosting capacity; see probabilistic hosting capacity. Scenario sampling must keep realistic combinations; see Monte Carlo load scenarios without impossible combinations. Cold-load pickup after outages adds a further synchronising event; see cold-load pickup after a long outage.
What this article does not cover
Coincidence factors for fast-charging stations are not covered here.
Where this reasoning stops. Probabilities are illustrative; local data on arrival and tariff response are needed.
Practical takeaway
Model EV charging as group behaviour with arrival patterns, tariffs and control failure, and size or manage transformers for synchronised scenarios rather than independent-car estimates.