Large-Scale EV Charging Using Genetic Algorithm with Greedy Constraint Repair Mechanism

In this paper, we propose an efficient approach based on genetic algorithms for scheduling the charging of large-scale electric vehicles (EVs) under various constraints. This problem is typically framed as an integer linear programming (ILP) issue, akin to other assignment problems. Two critical and unresolved challenges in EV charging are minimizing the total charging cost (Objective 1) and meeting both customer and grid constraints (Objective 2). To address Objective 1, we introduce the fitness function defined in Eq (3) to evaluate charging costs. The population evolves towards minimizing these costs until the termination condition is met. For Objective 2, we propose a swap operation that ensures time and state-of-charge (SOC) constraints are maintained throughout the repair process. Additionally, we present customized repair operators to address power limit constraints and three-phase imbalance constraints. In comparison to two widely used solvers, CBC and Surro-gateopt, our method demonstrates superior performance. CBC struggles with more than 1500 EVs, leading to extended computation times, while Surrogateopt encounters resource limitations with over 50 EVs. Our approach, however, is capable of finding near-optimal solutions for larger-scale EV problems on the same computer, with greater efficiency than these solvers. Moreover, our proposed algorithm exhibits nearly linear computational complexity, enabling it to handle large-scale constraint problems effectively.