Port of Rotterdam – truck electrification optimiser (v2) - by infrAi GmbH (Author: Mathias Schmidt)
study area: Maasvlakte · Europoort · Botlek · Pernis · Vondelingenplaat · Eem-/Waalhaven · Merwe-Vierhavens (generous extent)
map data: © OpenStreetMap contributors (ODbL), extract: Geofabrik Zuid-Holland 2026-11-16
No liability is assumed for the accuracy of the data, calculations, or representations
Loading port data …
—
—
Optimisation result vs. status quo
Model & assumptions

Executive logic

The tool sizes a truck fleet that moves road containers from the container terminals of the Port of Rotterdam to handover points (port exits, distriparks, empty depots, rail/barge hubs, customs) and decides, at the same time, which charging infrastructure to build: static charging sites, depot chargers and dynamically electrified road sections (ERS), each connected to an OSM substation with an assumed free capacity. One mixed-integer linear programme (MILP) minimises the annualised total cost of ownership (capex annuities + opex). The diesel-only status quo serving the same demand is solved as reference. A detailed German write-up of the formulation is delivered as PDF (MILP_Formulierung_Hafen_Rotterdam.pdf).

1. Demand and lanes

Per terminal t the user enters vessel calls per week, container moves per call, transshipment share and road share. Each road container is one loaded leg; with a backhaul share β a round trip carries 1+β containers.

Road containers_t = calls_t × moves_t × (1 − transshipment_t) × road share_t Round trips_t = Road containers_t / (1 + β) R_l (lane t→h) = Round trips_t × w_h / Σ w (w_h = weight of handover point h)

Each lane uses the fastest path on the OSM truck network (gate → handover and back, oneway-aware, speed = min(maxspeed, class default) × congestion factor). Round-trip time per vehicle type: driving × (1 + break/driving-limit) + handling at both ends (+ refuelling for diesel).

2. Decision variables

  • x(l,k) round trips/week on lane l by vehicle type k; N(k) vehicles of type k (integer)
  • y(σ) electrify road section σ (binary); v(g) ERS feed-in at substation g (binary)
  • z(s) build charging site s (binary); n(s) chargers at s (integer)
  • f(l,k,s) energy charged during handling at the lane's own endpoint s; qx(k,s) extra charging that costs driver time
  • w(l,k,σ) energy picked up dynamically on section σ; qd(k,d), nd(k,d) depot energy and vehicles at depot d

3. Constraints (weekly operation)

Demand: Σ_k x(l,k) = R_l ∀ l Energy (BEV): Σ_l E(l,k)·x(l,k) = Σ_{l,σ} w + Σ_{l,s} f + Σ_s qx + Σ_d qd ∀ k Per trip range: (E(l,k) − B_k)·x(l,k) ≤ Σ_σ w(l,k,σ) + Σ_s f(l,k,s) ∀ l,k (if E > B) Dynamic: w(l,k,σ) ≤ P_pick,k · η_ERS · h(l,σ) · x(l,k); Σ w(·,·,σ) ≤ M_σ · y(σ) Handling time: f(l,k,s) ≤ t_handling,s · p(k,s) · x(l,k) Chargers: Σ (f + qx)/p(k,s) ≤ n(s) · H_op · u_max ; n(s) ≤ n_max,s · z(s) Depot: qd(k,d) ≤ min(P_dep·η·H_off ; B_k · off-periods) · nd(k,d); Σ_d nd(k,d) ≤ N(k) Fleet hours: Σ_l T(l,k)·x(l,k) + Σ_s τ(k,s)·qx(k,s) ≤ N(k) · H_op · availability / peak factor Grid: Σ_{s→g} sim·P_s·n(s) + Σ_{σ→g} P_ERS,σ·y(σ) + Σ_{d→g} sim·P_dep·nd ≤ Cap_g ∀ g ERS feed: y(σ) ≤ v(g(σ))

E(l,k) = round-trip energy, B_k = usable battery (SoC window, constant-power range ≤ 80 %), h(l,σ) = hours per round trip on σ, p(k,s) = min(charger, vehicle) power × η, τ = hours per kWh incl. plug-in time.

4. Objective (€/yr)

min Σ_k N_k·(capex_k·CRF + fixed_k + depot charger) + Σ_σ y_σ·lane-km_σ·(capex_ERS·CRF + O&M) + Σ_g v_g·(connection + cable)·CRF + Σ_s [ z_s·(site + connection + cable)·CRF + n_s·charger·CRF ] + weeks · [ Σ x·(km·maintenance + T·wage + diesel·(price + CO₂)) + Σ energy·(price + CO₂)/η + Σ qx·τ·wage ]

5. Solution procedure

HiGHS (WebAssembly) solves the MILP in the browser within the time limit and gap. If the limit is reached, the best road-section layout is fixed and fleet, sites, chargers and grid connections are re-optimised (polish step). Road sections are formed from candidate OSM edges that are used by exactly the same lanes and fed by the same substation (≤ 2 km each).

6. Key simplifications – please review

  • Weekly aggregation with a peak factor instead of a time-resolved vessel schedule (container dwell times smooth the vessel peaks).
  • Fixed fastest routes; no route choice towards electrified roads; nearest eligible substation per asset.
  • Free substation capacity is not in OSM – class defaults are assumptions and must be replaced with grid operator data (Stedin / TenneT capacity maps).
  • Port exits are treated as handover hubs (shuttle concept). Hinterland trucks that leave the port are not sized by this model.
  • Terminal-internal equipment (AGVs, straddle carriers) is out of scope; RWG has no OSM polygon (manual gate position).
  • Single representative year; vintages, technology cost trends and autonomous driving (as in the infrAi Fleet Optimizer) are not yet included.
Optimising … (HiGHS)