基于不确定需求的集中式充电站分级选址鲁棒优化研究

Robust Optimization of Hierarchical Site Selection Location of Centralized Charging Station Based on Uncertain Demand

以集中式充电站为研究对象,采用多面体不确定集描述不同用地性质下异质充电需求的不确定性,考虑运营商和用户双方利益,构建以建设运营成本、充电成本和寻站成本最优为目标,充电距离和充电需求等约束协同的鲁棒优化模型,并通过Bertsimas和Sim鲁棒对等转换,将半无限优化模型转换成易于求解的0~1整数规划问题。数值仿真结果表明:首先,集中式充电站选址不确定水平因素发生变化时,系统成本呈现阶梯式上升,并且建设运营成本对系统成本影响最为显著;其次,充电需求变化幅度越小,模型鲁棒性越高,集中充电设施选址适应度越高,充电站建设运营商根据自身的风险偏好程度选择不同参数组合,实现在异质需求下集中式充电站充电需求与社会总成本的平衡,获得充电站选址的科学方案;最后,对比分析不同情景下的充电站设置方案验证模型有效性。

Centralized charging stations were taken as the research object.A polyhedral uncertainty set was used to describe the uncertainty of heterogeneous charging demands under different land use properties.Considering the interests of both operators and users,a robust optimization model was established with the goal of optimizing construction and operation costs,charging costs,and station-finding costs,and with the coordinating constraints such as charging distance and charging demand.Bertsimas and Sim robust equivalence transformation was used to convert semi-infinite optimization model into an easily solvable 0~1 integer programming problem.Numerical simulation results indicate that:firstly,as the uncertainty level factor for centralized charging station location changes,the system costs exhibit a stepwise increase,and the construction and operation costs have the most significant impact on system costs.Secondly,the smaller the variation range of charging demand,the higher the robustness of the model,and the higher the adaptability of centralized charging facility location selection.Charging station construction operators can select different parameter combinations according to their risk preferences,achieving a balance between the charging demand and the overall societal cost of centralized charging stations under heterogeneous demand,thereby,scientific plan for charging station site selection is obtained.Finally,charging station configuration schemes in various scenarios are compared and analyzed and the effectiveness of the proposed model is verified.