The impact of rising battery-electric vehicle (BEV) penetration on low- and medium-voltage distribution grids (LV/MV), and the extent to which distribution system operators (DSOs) can manage emerging congestion through grid-fee modulation or load reduction, remains an unresolved issue. The proposed co-simulation framework couples charging-activity demand from large-scale agent-based transport simulation (MATSim) with power flow analysis (pandapower). MATSim simulates realistic human behaviour through individual agents with daily activity chains, so that charging demand emerges endogenously from where and when agents park. Pandapower then takes these charging events as load demand and runs power flow on the synthetic LV/MV topology. A time- or location-varying grid fee only works if agents respond by shifting charging time, location, or trip timing. The coupled framework captures this feedback: the fee signal is changed, MATSim re-optimises agent behaviour, and the grid response is re-evaluated. A pure power-system model would have to assume the demand response, while a pure transport model could not verify whether the grid problem was resolved.

Standard load profiles represent non-EV baseline demand, while EV load is generated endogenously from simulated travel, parking, and charging decisions. The study evaluates a future scenario with 100 % BEV penetration of private vehicles (650,000 EVs), based on the Open Hamburg scenario. Existing charging infrastructure is complemented by planned public and private stations, positioned according to social, economic, and geographical factors.

The methodological contribution lies in integrating three elements often treated separately: BEV charging demand, feeder-level grid constraints, and tariff-sensitive driver behaviour. The framework is applied to a synthetic LV/MV distribution grid of Hamburg generated with pylovo, providing a realistic feeder topology for evaluating localised congestion. At its core is a DSO control layer that modulates a congestion-dependent grid fee and/or load reduction when local voltage or loading thresholds are approached, steering charging away from critical states. Driver response is modelled through a utility score with low, medium, and high price-sensitivity tiers. Grid-side metrics include voltage violations, line and transformer loading, losses, and peak demand, user-side metrics include shifted energy and charging cost.

The results are threefold. Uncontrolled charging intensifies evening peaks and creates localised bottlenecks. Smart charging reduces peak demand and technical violations. Grid-fee modulation induces measurable load shifting across all driver tiers. With a maximum fee of 1 €/kWh and charge-point power capped at 4.2 kW, transformer loading drops from 300 % to 100 % and minimum voltage rises from 0.90 to 0.95 p.u. With further optimization (e. g. AI) is sufficient to resolve the congestion observed under uncontrolled charging, without reinforcement of the physical grid.