Existing electric vehicle (EV) scheduling schemes fail to address the issue of balanced utilization of charging piles, which often leads to overloading and premature aging of certain charging piles while others remain underutilized. Concurrently, vehicle-to-grid (V2G) technology enables bidirectional energy flow, enhancing grid regulation capabilities while providing users with discharge revenue. In light of this, a two-stage EV scheduling optimization method based on charging pile allocation and charge-discharge scheduling is proposed. In the first stage, the allocation of charging piles is optimized with the objective of minimizing the variance in charging pile utilization. In the second stage, the charge-discharge power of EVs is optimized to achieve threefold objectives: minimizing the variance of regional load, minimizing user charging costs, and maximizing charging pile revenue, thereby achieving tripartite collaborative optimization. An adaptive genetic algorithm (AGA) is employed to solve the established bi-level EV scheduling model. The case study results demonstrate that, compared to conventional strategies that neither consider charging pile load balancing nor incorporate V2G technology, the proposed method reduces the variance of charging pile utilization by 93.6%, decreases the variance of transformer area load by 16.5%, lowers users' net charging costs by 12.0%, and increases the charging station's daily revenue by 14.4%. These outcomes fully substantiate the method's superior performance in optimizing charging infrastructure utilization, mitigating load fluctuations, and enhancing multi-stakeholder economic benefits.