A Review on Optimizing Electric Vehicle Charging Schedules for Cost and Grid Efficiency

Abstract

The rapid growth of electric vehicles has introduced new operational and economic challenges for modern power systems, including charging demand management, grid stability, peak load reduction, and cost control. To address these issues, various optimization strategies have been proposed, including heuristic and metaheuristic algorithms, exact mathematical programming models, and data-driven approaches. Heuristic and metaheuristic methods such as Genetic Algorithms, Particle Swarm Optimization, and evolutionary algorithms are widely used because of their flexibility and ease of implementation. However, they may suffer from parameter sensitivity, limited reproducibility, and the absence of guaranteed optimality. In contrast, exact optimization approaches, particularly Mixed-Integer Linear Programming, provide a structured framework for modeling constraints such as charging power limits, state-of-charge requirements, transformer capacity, dynamic pricing, and vehicle-to-grid interactions. This review examines studies published after 2020 and compares solver-based, heuristic/metaheuristic, and data-driven approaches for electric vehicle charging scheduling in terms of cost efficiency, grid impact, scalability, constraint handling, and practical applicability. Rather than presenting new computational experiments, the paper synthesizes reported findings from the literature and discusses the conditions under which solver-based methods, including IBM CPLEX, are advantageous or limited. The reviewed studies suggest that solver-based methods are useful for constraint-rich and reproducible scheduling formulations, while heuristic and data-driven methods remain attractive for large-scale, uncertain, or real-time applications. The review highlights unresolved gaps, including limited standardized benchmarks, insufficient empirical comparison across common test systems, weak integration of battery degradation and user behavior, and a lack of realistic uncertainty and grid-aware modeling.