Fairness and coverage optimization in electric vehicle charging networks with exact and reinforcement learning-embedded solution methods

The widespread adoption of electric vehicles (EVs) requires analytical tools to guide sustainable charging infrastructure planning. This study proposes a bi-objective multi-period mixed-integer nonlinear programming (MINLP) model to optimize EV charging networks by maximizing coverage and fairness under budgetary constraints. Coverage ensures that EVs complete trips without depleting their charge, while fairness guarantees equitable access with impartial charging times. To solve the model, we employ the augmented ε-constraint method and a reinforcement learning-embedded non-dominated sorting genetic algorithm (NSGA-II). Integrating reinforcement learning into the optimization framework significantly enhances the model’s scalability and adaptability, enabling the efficient solution of large-scale road network charging station planning problems. Using a real road network dataset, we analyze the effects of EV range, state of charge (SoC), and budget allocation. Results show that enhancing fairness is possible with only minor reductions in coverage. Moreover, SoC at destinations exerts greater influence than SoC at origins, while extending EV range leads to significant budget savings when a higher destination SoC is required. These findings offer practical insights for planners and policymakers to design fair, efficient, and sustainable EV charging infrastructures.