Robust Optimisation for Electric Vehicle Charging Station Siting Under Demand Uncertainty: A Multi-Station Network Planning Model

The accelerating adoption of electric vehicles (EVs) demands large-scale public charging infrastructure, yet planners must make long-horizon siting decisions under substantial uncertainty about future demand. This paper presents a mixed-integer linear programming (MILP) model for the robust siting of multi-station EV charging networks that explicitly handles demand uncertainty through the Bertsimas–Sim budgeted uncertainty framework (Bertsimas and Sim 2004). The model jointly optimises the selection of candidate sites, fractional demand allocation across stations, and reserve capacity for worst-case demand deviations, subject to distance-feasibility and capacity constraints. A case study using publicly available EV registration and charging-activity data from a high-penetration region demonstrates the practical applicability of the approach. Results show that geographic reach, rather than capital budget, is the binding constraint on coverage: opening all 482 economically viable candidate sites achieves only 67.4% demand coverage due to a binding 20 km service-distance limit. Scenario analysis reveals that a robustness budget of Γ = 30% of demand zones provides the most cost-effective protection, reducing coverage variability by 38% and capacity violation rates from 24.7% to 10.7% at an 8.5% cost premium over deterministic planning. The findings provide planners with a decision-support framework that is both computationally tractable and robust to realistic demand fluctuations.