Coordination, not energy volume, governs the transformer limit

What the model tests

A representative 75-kVA pad-mount service transformer serves a ten-vehicle depot. Each vehicle needs 40 kWh overnight on an 11.5-kW Level-2 charger; vehicles arrive between 17:00 and 19:00 and leave between 05:00 and 07:00, giving an overnight dwell. Four scenarios deliver identical total energy and differ only in when that energy is drawn: site base load alone (S1), unmanaged charging that begins at plug-in (S2), thermally-managed charging that spreads minimum power across the dwell (S3), and managed charging paired with a 30-kW / 120-kWh battery (S4).

The transformer's operating envelope is evaluated with the IEEE Std C57.91-2011 dynamic loading model, which relates per-unit load and ambient temperature to top-oil and hot-spot temperature and, through an Arrhenius relationship, to equivalent insulation loss-of-life. Because every scenario delivers the same energy, any difference in thermal stress is attributable to coordination, not consumption.

The finding

Unmanaged charging drives the transformer to 170.7% of nameplate and holds it above rated load for 3.75 hours a night. The hot-spot peaks at 161.4 °C, far past the 110 °C normal-aging reference, advancing equivalent insulation aging to roughly 8,373× the baseline rate, about five days of transformer life consumed per calendar day. This scenario is a deliberately severe bound: near-simultaneous plug-in, no diversity beyond the arrival window, and a flat 30 °C ambient. It marks the upper edge of harm, not a typical day.

Scheduling the same energy against the thermal limit cuts the peak by 61%, from 170.7% to 66.3% of nameplate, and eliminates rated-load overload entirely. The hot-spot falls to 71.0 °C and daily loss-of-life drops to about 9× baseline, elevated relative to a no-charging day, but well inside a sustainable aging envelope. Valley-filling alone restores full thermal headroom at this fleet-to-transformer ratio.

Adding storage changes nothing at this depot size. Because managed charging already holds loading below the battery's dispatch setpoint, the battery never sees a residual peak to shave. That is itself the result: where thermal-aware scheduling absorbs the coincident load, storage is optional until concentration rises.

When storage starts to matter

Storage is not universally required: it becomes decisive at a specific concentration. Holding every other input fixed and increasing only fleet size: at 15 vehicles, managed scheduling peaks at 91.5% and the battery trims it to 90.0%, with both cases still under nameplate. At 20 vehicles the behavior changes qualitatively. Scheduling alone can no longer hold the limit: it reaches 115.7% and 10.5 hours a day of overload, tripping the screen, while adding storage restores 90.0% loading, removes the overload entirely, and cuts daily loss-of-life from about 1,208× to 162× baseline.

The sequence is the point: schedule first, add storage when coincident load outgrows what scheduling can absorb. Storage earns its place precisely where thermal-aware coordination can no longer hold the asset on its own.

Why it matters

The depot transformer is the smallest fully computable version of the grid's defining problem: concentrated new load arriving faster than the infrastructure built to serve it. At the distribution edge, the binding constraint is transformer hot-spot temperature. At the transmission edge, it is the time-to-power gap between an approved project and an energized one. Recent FERC large-load interconnection activity contemplates expedited study for loads that agree to be flexible and curtailable: the same lever, one layer up.

A data center, an electrified plant, and an EV depot differ in magnitude but share one structure: a concentrated load, a constrained asset, a physical limit, and a flexibility resource. Coordinated flexibility relieves the constraint and defers capital build, provided coordination is measured against the asset, not the price of energy. The depot case is a distribution-level analogue of the large-load time-to-power problem, computed end to end.