Where electricity is cheaper at some hours than others, moving charging into the cheap window is an obvious saving and a straightforward feature to build.
It becomes interesting at scale, because a naive implementation creates a new peak at the moment the cheap period begins, and that peak can cost more than the energy saved.
The synchronised start problem
If every charger begins at the instant the cheap rate starts, the site steps from near zero to full charging load in one moment. On a site with a demand-based tariff component, that peak may erase the saving entirely.
Staggering starts across the first part of the window, with randomisation so that units do not cluster, spreads the load. The saving is preserved and the peak is not created.
Charge to a target, not for a duration
The useful specification is a state of charge by a departure time rather than a fixed charging window. That lets the system distribute charging across whichever hours are cheapest while still meeting the requirement.
It requires knowing the departure time, which for a fleet is known and for a public site is not. This is why tariff-aware scheduling works well in depots and poorly at destinations.
Guarantee the outcome, optimise within it
A scheduler that saves money and occasionally leaves a vehicle short will be switched off after the first occurrence. The requirement is to meet the charge target reliably and optimise cost within that constraint.
That means falling back to expensive hours when necessary rather than treating the price signal as absolute, and it means being conservative about how much cheap time is assumed available.
Vehicles have their own opinions
Many vehicles have their own scheduling, set by the driver and invisible to the charger. A vehicle that has decided to start at a particular time will not draw before it, regardless of what the charger offers.
Two schedulers with different intentions produce sessions that do not happen. Where a site controls charging centrally, drivers should be told to disable vehicle-side scheduling, and where they cannot be told, the system must tolerate it.
Interaction with load management
Tariff scheduling wants to concentrate charging into cheap hours. Load management limits how much can happen at once. Together they define a smaller feasible region than either alone, and on a constrained site the cheap window may not be long enough for every vehicle.
Prioritisation then decides who charges cheaply and who does not, which is a business rule rather than an optimisation.
Confirm the tariff actually varies
Not every connection is on a time-varying tariff, and some that are have differentials too small to be worth engineering around. Establish the actual structure and differential before building scheduling that assumes one.
Where a demand component exists, it usually dominates, and managing peak matters more than shifting energy.