Charging infrastructure for last-mile delivery vans

The vans come back to the same yard every night, so there is somewhere to build. The routes are short enough that daily energy is modest, so the ports do not have to be powerful. The dwell is long, so there are hours to spread that energy across. And the fleet is large and uniform, so the arithmetic is stable and the per-port cost of a big build lands at the bottom of the range instead of the top.Uniformity is the underrated one. A mixed fleet has to be sized for its hardest case; a yard of identical vans running comparable routes can be sized on a distribution rather than a worst case, and managed charging can treat them as interchangeable. That is what makes shared ports work here when they struggle elsewhere.The consequence is that the design conversation should start at managed Level 2 and only move up if the arithmetic pushes it there. DC fast charging at a last-mile depot is usually a small number of positions for exception cases — a van back short, a mid-shift turnaround, a rental brought in at peak — rather than the backbone of the site. For reference, national installed cost for depot Level 2 runs roughly $3,500 to $15,000 per port, and DC fast starts an order of magnitude above that.

In an ordinary month that ratio is commonly well under one half, which is why a forty-van depot does not need forty ports. What matters is that the ratio is not a constant. It moves with three inputs, and all three move against you at the same time of year: the route gets longer, the van gets less efficient in the cold, and the dwell gets shorter because the vans go out earlier and come back later.Run the table below with your own numbers and the design decision falls out of it. The row you build to is not the first row. It is the last one.

This is the second-order consequence of the table above, and it catches electrical designs rather than operations. In an ordinary month a forty-port row draws an average power far below its nameplate, so the controller has enormous freedom to shape the load, keep the site inside the existing service, and hold the demand peak down. At peak season in cold weather the same row has to deliver nearly all of its nameplate energy in a shorter window, and there is almost nothing left to manage.So the site can be inside its service on every ordinary day of the year and outside it on the days that matter most. Size the service against the peak row, not the average one, or accept in advance that peak-season vans will leave partially charged and plan the routes around it.It also means the demand charge you see in December is not the one you budgeted from a spring bill. If your tariff has any kind of ratchet provision, one peak-season month can affect billing for a long time afterwards. Read the schedule, and ask the utility what the ratchet does before the first peak season rather than after it.

Routes lengthen, so each van needs more energy. Dispatch moves earlier and returns move later, so the dwell window shrinks. And in most of the country it is cold, so energy per mile rises. Multiply those together and the port requirement can roughly triple against an ordinary month, which is exactly the sort of change a design sized on an annual average will not absorb.Then the fourth thing happens: extra vans. Seasonal capacity arrives as rentals or contracted vehicles, they are often not the same model, and nobody asked at the time of hire whether the yard had ports for them. A depot that was tight at forty vans is now being asked to charge fifty-five.The result is a queue at eleven at night, a decision about which vans get charged, and a set of routes going out short. None of that is an equipment fault and none of it will be fixed by the vendor. It is a sizing decision made in July.The honest mitigation list is short. Build the service and the conduit for the peak case. Populate ports up to the busy-month case and hold a plan for adding the rest. Book seasonal vehicles with charging in the requirement rather than as an afterthought. And know in advance which routes get shortened if the yard cannot deliver, so the decision is made by a plan rather than by whoever is on the yard at midnight.

The civil work is the part you cannot cheaply redo, and reopening a working yard is disruptive as well as expensive. Conduit is also the cheapest thing in the trench, so oversizing it is the least regrettable decision on the whole project. Distribution equipment sits in the middle: expensive enough to think about, and much harder to add later than to include now.The powered equipment is the part that dates, so it is the right thing to defer. Chargers improve, connector standards move, and a port you did not buy in year one is a port you buy cheaper and better in year three. Split the project at that line: passive infrastructure for the eventual build, powered equipment for the next two years.Talk to the utility before any of this is fixed. The service capacity at your yard, the cost to change it and the lead time on the equipment they would need to install are the three answers that reorder every other decision, and they have the longest calendar. On money, there is no federal charger credit to net against the build — Section 30C terminated for property placed in service after June 30, 2026 — so anything recoverable will be a utility make-ready program or a state program, both sponsor-specific, and both to be read on the sponsor's own page with the date you read it.

Energy per mile from telematics, monthly, gives you your own derate instead of somebody else's assumption. Route miles by month gives you the real seasonal profile rather than the planned one. Actual dwell — measured from wheels-stopped to wheels-moving, not from the shift roster — is almost always shorter than the roster implies, and it is the input the design is most sensitive to. And plug-in compliance tells you how much of any shortfall is process rather than capacity, which is the difference between buying ports and fixing a workflow.Four numbers, all of which you already generate, none of which most depots retain. Put them in a monthly report and the next capital request writes itself with your own data in it, which is a much stronger paper than a vendor's model.