The SoC risk problem that keeps fleet managers awake at 05:00
Fleet electrification introduces an operational dependency that diesel fleets never faced: every vehicle has a finite energy budget expressed as battery state of charge (SoC), and that budget must be sufficient for the day's planned route before first departure. A diesel truck that enters the yard with a quarter tank can be topped off in 5 minutes. A BEV rigid truck that enters the depot at 15% SoC at 19:00 and is scheduled for a 05:30 departure may require 10–12 hours at 22 kW AC to reach 95% SoC — exactly the window available overnight. If anything disrupts that charging window (charger fault, connection drop, CPMS communication failure, power curtailment by the DSO), the vehicle exits the depot undercharged and the day's route must be shortened, re-planned, or the vehicle pulled from service.
The consequence of undercharging is not merely inconvenient. For cold-chain logistics operators running temperature-controlled deliveries on fixed route schedules, a vehicle that can only cover 70% of its planned route before requiring a mid-route charge stop creates a cascade of delays across the delivery manifest, potentially triggering SLA penalty clauses with retail customers. The operational cost of a single morning departure event where 3–5 vehicles are undercharged can run €2,000–5,000 in schedule disruption, re-routing, and customer communication costs.
This is why the departure readiness guarantee is, for fleet managers, the non-negotiable first requirement of any smart charging system. Cost optimisation is valuable. Demand charge management is valuable. But neither matters if vehicles miss departures or exit undercharged. The constraint ordering is explicit: departure SoC requirements are hard constraints; energy cost and demand charges are optimisation objectives within the feasible space defined by those hard constraints.
Modelling departure readiness as a hard constraint
In the MILP scheduling formulation, departure readiness is expressed as a constraint on the state-of-charge at the departure time slot for each vehicle. For vehicle i with departure time t_dep(i) and target SoC SoC_target(i):
SoC(i, t_dep(i)) ≥ SoC_target(i)
This constraint must be satisfied regardless of the objective function's preference for cheap energy hours. If the only feasible way to deliver SoC_target(i) to vehicle i by t_dep(i) is to charge at maximum current during the peak tariff window from 18:00 to 22:00, then that's what the scheduler must do — and it should flag this as a high-cost session that exceeded the tariff optimisation budget, for post-hoc analysis.
The practical implementation adds a safety margin: target SoC is set to SoC_target + buffer, where the buffer accounts for BMS SoC measurement uncertainty (typically ±3–5% for lithium iron phosphate (LFP) cells, ±2–3% for NMC chemistries), energy losses in the AC charging circuit, and the residual discharge during overnight parking in cold temperatures (typically 0.5–1% SoC loss per hour at 0°C ambient for well-insulated battery packs). For a vehicle with a 95% required departure SoC and 5°C overnight temperature, a realistic scheduling target is 97–98% to absorb the overnight thermal drain before the driver connects the vehicle to the route planning computer at 05:15.
The departure window and why it's not a single timestamp
Most fleet management systems store departure times as planned schedule entries, but actual departure events have variance. A driver might connect to their vehicle 20 minutes before planned departure to complete pre-trip inspection and cabin warm-up. Another vehicle might be held at the dock for an unexpected late-arriving pallet and depart 45 minutes after its scheduled window. The scheduler needs to handle this variance without either over-charging (wasting energy) or under-charging (failing the readiness constraint).
The reliable approach is to model each vehicle's departure as a window rather than a point in time: [t_dep_earliest(i), t_dep_latest(i)]. The scheduler must ensure that SoC_target(i) is achieved by t_dep_earliest(i) — the earliest plausible departure — and that charging continues at maintenance level (typically 80–95% SoC range, avoiding full charge hold which degrades LFP cycle life) until t_dep_latest(i) if the vehicle remains connected. This window-based approach absorbs the typical ±45 minute departure variance seen in logistics operations without either under-delivering energy or over-cycling batteries.
The departure window data usually comes from the fleet management system (FMS), not from the CPMS. Integration between FMS departure schedules and the charging management system is one of the most important — and most frequently missing — links in fleet electrification deployments. Depots that run fleet scheduling in a separate tool from charging management typically lack this integration and resort to manually configuring departure times in the CPMS each morning, which introduces human error and adds operational overhead.
A scenario: mixed-departure depot, 38 vehicles, three departure waves
A parcel delivery depot near Cologne operates 38 BEV light commercial vehicles with three departure waves: an early wave at 05:30 (12 vehicles on metropolitan routes), a mid-morning wave at 07:00 (18 vehicles on suburban routes), and a daytime wave at 09:30 (8 vehicles on business-district deliveries). Vehicles return between 17:00 and 20:30. The depot has a 250 kW EV charging budget from 18:00 onwards after baseline building loads are accounted for.
The scheduling problem has a clear priority ordering: the 12 early-wave vehicles have the least time to charge (at most 11.5 hours from 18:00 to 05:30) and must be served first in any priority conflict. The 18 mid-wave vehicles have 13 hours. The 8 daytime vehicles have 15.5 hours — the most buffer. The scheduler assigns priority tiers and enforces them via per-vehicle TxProfile overrides for vehicles in the early wave that arrive with critically low SoC (<25%), while the mid-wave and daytime vehicles are charged on TxDefaultProfile schedules weighted toward the cheap 22:00–04:00 tariff window.
On a typical Wednesday: all 12 early-wave vehicles arrive between 17:00 and 19:00 with average SoC of 28%. The scheduler calculates that 9 of the 12 have sufficient time on the overnight cheap window alone; the remaining 3 arrived with <20% SoC and require immediate high-power charging from 18:00 to ensure readiness by 05:30. The scheduler assigns those 3 vehicles TxProfile overrides at 32A (22 kW) from the moment of connection. Their Arbeitspreis cost for the session is approximately 35% higher than the overnight optimised sessions. The daily cost premium for those 3 emergency sessions is roughly €4.50 above what tariff-optimal scheduling would have achieved. Fleet managers consider this an acceptable cost for guaranteed departure readiness on those vehicles.
Failure modes: where departure guarantees break down
We're not saying that smart charging software can guarantee departure readiness under all conditions. Several failure modes can defeat even a well-designed scheduling system:
OCPP session continuity failures. If a charge point firmware crashes and restarts mid-session, the active TxProfile assigned to an in-progress transaction may be lost (depending on whether the firmware persists in-memory session state to flash on restart). The CPMS must detect the BootNotification / transaction reconciliation and re-push the necessary profiles immediately. Any delay in this recovery adds time at zero or minimum charging rate to a vehicle that may already be on a tight schedule.
DSO §14a curtailment during overnight window. German DSOs exercising their §14a EnWG steuerbare Verbrauchseinrichtungen rights can curtail EV charging loads during grid congestion. A 2023 BNetzA framework decision permits curtailment to 4.2 kW per connection point. For an overnight charging window that relied on 22 kW per vehicle, a 4.2 kW DSO curtailment signal reduces energy delivery to 19% of planned — vehicles that were on track for 95% SoC by 05:30 will arrive at departure at 40–50% SoC instead. Depots in areas with congested low-voltage distribution networks need to plan for this scenario, either through oversized overnight windows that absorb curtailment events, or through battery storage buffers that charge at medium-priority times and discharge to EV charging during DSO curtailment periods.
Vehicle BMS communication failures. The scheduling engine's SoC model depends on accurate BMS data from the vehicle, typically polled via the CPMS MeterValues messages with the SoC measurand. If a vehicle's BMS reports SoC incorrectly (a known issue with some LFP batteries in cold temperatures where the cell voltage-SoC curve is nearly flat below 20°C, making estimation unreliable), the scheduler may believe the vehicle is more charged than it actually is, and under-prioritise it. A conservative scheduling approach buffers against this by using the reported SoC minus 5–8% as the planning baseline for any vehicle operating in sub-10°C conditions.
Metrics that confirm the guarantee is working
A fleet operator running a smart charging deployment that claims to guarantee departure readiness should monitor three metrics weekly:
Departure readiness rate: percentage of planned departures where actual vehicle SoC at departure time met or exceeded the target SoC. Target: ≥99.5%. Any week below 98% warrants root-cause analysis.
Emergency override frequency: percentage of charging sessions requiring a TxProfile emergency override (i.e., where the default overnight schedule was insufficient for the vehicle's departure requirement). A rising trend indicates either fleet utilisation increasing beyond scheduling assumptions, or vehicles arriving with consistently lower SoC than planned — which points to a route energy model that needs updating.
Average departure SoC buffer: the mean (actual departure SoC − target SoC) across all vehicles. Consistently high buffers (e.g., average +12% above target) indicate that the scheduling is over-conservative, charging vehicles earlier than necessary and potentially incurring higher tariff costs. Consistently low buffers (average +2% above target) indicate the schedule is tight and any disruption will immediately produce readiness failures. The target buffer depends on fleet operational variance, but 5–8% average margin is typical for well-tuned depot schedules.