Load balancing is usually specified as a yes or no requirement and is really a choice between three approaches with different infrastructure needs and different failure behaviour.
The distinguishing question is not how well each performs when everything works. It is what each does when the controller, the meter or the network is unavailable.
Static allocation
Each charger is configured with a fixed limit chosen so that all bays at their limit stay within the supply. Nothing measures anything and nothing communicates.
It is robust, needs no infrastructure, and wastes capacity whenever the site is not full, which is most of the time. On a site with occasional peak occupancy it can be entirely adequate.
Dynamic allocation
A controller measures actual site load and distributes the remaining capacity among active chargers, adjusting as vehicles connect and finish. This recovers the capacity static allocation leaves unused.
It requires a meter at the supply point, a controller, and a communication path to every charger. Each of those is a dependency, and the design has to define what happens when any of them is lost.
Decentralized allocation
Chargers negotiate among themselves over a local link rather than being directed by a central controller. There is no single point whose failure stops allocation, and the system continues working when the backend is unreachable.
It is harder to reason about, harder to observe, and the emergent behaviour under unusual conditions is less predictable. It suits sites where connectivity is unreliable and continuity matters more than central visibility.
Compare them on failure, not on performance
This comparison, rather than peak utilisation, is what should drive the choice on a site where a trip affects more than the chargers.
- Static: nothing to fail. Capacity is permanently under-used.
- Dynamic with cloud control: loses allocation when connectivity drops. Fallback limit must be safe.
- Dynamic with a local controller: survives internet loss, fails if the controller does.
- Decentralized: survives both, provided the local link between chargers holds.
The meter is the dependency people forget
Dynamic allocation depends on knowing the non-charging load. If that measurement is wrong, stale or absent, the controller is either over-allocating, which is unsafe, or under-allocating, which wastes the capability it was installed for.
Meter placement and reliability therefore deserve as much attention as the chargers, and a site without a suitable measurement point cannot do true dynamic allocation regardless of what the chargers support.
Mixed sites
Large sites frequently end up with a hierarchy: a site-level budget divided among groups, with allocation within each group. That is more complex and it maps better onto real electrical topology, where sub-boards have their own limits.
Designing the allocation hierarchy to match the electrical hierarchy avoids a system that respects the site total while overloading a distribution board.