近年来，中国乘用车市场的消费结构正经历着深刻变革。电动汽车凭借其低碳环保、充电成本低廉等优势愈受青睐，电动汽车在新车总销量占比已突破30%，作为其协同配套的充换电市场亦呈现高速增长态势。在电动汽车及其补能设施规模快速发展的背景下，大规模充换电设施集群有潜力成为新型灵活性资源参与电网调控。如何对大规模电动汽车充放电功率进行调控、对充放电需求的时空分布进行价格引导，成为亟待解决的现实性问题。本文从电动汽车充换电运营商视角，面向功率调控和价格引导两种调控手段，系统性地研究了以下四方面问题：（1）针对大规模补能设施功率调控灵活性评估问题，使用集总累积能量-功率边界对充电站和换电站的功率调控灵活性进行统一建模；针对换电站的储备运行方式，提出考虑电池储备的灵活性评估方法。考虑集总边界模型在逐日决策中的累积能量状态耦合，构建端点形集总累积能量-功率边界模型，并进一步提出双层遍历方法以得到灵活性损失最小的端点形集总边界。（2）针对大规模补能设施功率调控灵活性预测问题，提出了基于长短期记忆网络的确定性预测方法以及基于条件变分自编码器的场景预测方法。在条件变分自编码器训练过程中，使用1-近邻分类模型和留一法交叉验证来评估场景生成效果，并指导模型超参数的调整。（3）针对大规模补能设施聚合调控问题，考虑运营商同时参与日前能量市场和实时能量市场，提出一种风险规避的投标策略。在得到日前市场出清的集总指导功率后，提出了考虑前瞻和再分配环节的充换电设施集总功率实时分解方法，可实现大规模充换电设施的充放电功率的快速、准确调控。（4）针对充电需求时空转移灵活性评估与价格引导问题，基于条件随机场模型对充电需求时空转移灵活性进行评估。面向公共充电站，提出了考虑充电需求时空转移的分时分区充电价格优化策略，可引导充电需求在时空上的合理分布。本文基于充换电运营商实际的充电和换电记录数据，对所提出的模型和算法进行了验证。本文的工作为充换电运营商聚合可调控的充换电资源参与电力市场或提供辅助服务提供了灵活性评估预测、投标决策和充放电功率调控的理论方法和有效工具。同时，公共充电站的定价策略可引导电动汽车充电负荷在时空上的合理分布。
In recent years, the consumer structure of China's passenger vehicle market is undergoing profound changes. Electric vehicles, favored for their low carbon footprint and cost-effective charging advantages, have seen their share in new car sales exceed 30%, with the charging and swapping market experiencing rapid growth. Against the backdrop of swift expansion in electric vehicles and their charging infrastructure, large-scale charging facilities clusters have the potential to become novel flexible resources for grid regulation. How to regulate the charging and discharging power of a large number of electric vehicles and guide the spatial and temporal distribution of charging and discharging demand through pricing has become a pressing issue. This dissertation, from the perspective of electric vehicle charging and swapping operators, systematically studies the following aspects concerning power control and price guidance as regulatory measures: (1) For the flexibility assessment of large-scale charging facilities' power regulation, a unified modeling approach using the Aggregate Cumulative Energy and Power Boundary Model is applied to assess the charging and swapping stations' discharge and charge power flexibility. Specifically, for battery swapping stations, a flexibility assessment method considering reserves is proposed. Considering the aggregated boundary model in daily decision-making involves coupling of energy states, an endpoint aggregated energy and power boundary model is constructed, and a dual-layer traversal method is further proposed to minimize flexibility loss. (2) Regarding the prediction of power regulation flexibility for large-scale charging infrastructure, a deterministic prediction method based on Long Short-Term Memory (LSTM) networks and a scenario prediction method based on Conditional Variational Autoencoders (CVAEs) are proposed. During the training process of CVAEs, a leave-one-out cross-validation with the 1-nearest neighbor classification model is used to evaluate the scenario generation performance and guide the adjustment of model hyperparameters. (3) For the issue of aggregate control of large-scale charging infrastructure, considering the operator's participation in both the day-ahead energy market and the real-time energy market, a risk-averse bidding strategy is proposed. After obtaining the cleared aggregate power from the day-ahead market, a real-time decomposition method for the aggregate power of charging and swapping facilities, considering foresight and redistribution, is proposed to achieve rapid and accurate regulation of charging and discharging power for large-scale facilities. (4) Regarding the issue of flexibility assessment and price guidance for spatial and temporal transfer of charging demand, the flexibility of spatial and temporal transfer of charging demand is assessed based on the Conditional Random Field (CRF) model. Aimed at public charging stations, a time-of-use and zone-specific charging price optimization strategy considering the flexibility of spatial and temporal transfer of charging demand is proposed to guide the reasonable transfer of charging demand in space and time.This dissertation validates the proposed models and algorithms based on actual charging and battery swapping record data from charging and swapping service operators. The work presented in this dissertation provides theoretical methods and effective tools for flexibility assessment and prediction, bidding decision-making, and charging and discharging power regulation for charging and swapping service operators to aggregate controllable charging and swapping resources for participation in the electricity market or providing auxiliary services. Moreover, the pricing strategy of public charging stations can guide the reasonable distribution of electric vehicle charging load in both time and space.
