伴随我国经济的快速发展和汽车工业的整体进步，我国机动车保有量逐年上升。但随之而来的是我国停车位的严重短缺，停车设施供需严重不平衡等问题。与此同时，我国停车市场的发展愈发迅猛，停车场的收益管理问题越发受到重视。在都市核心区的停车场运营中，客户群体相对固定，因此停车场推出了停车月卡产品以满足客户需求。但大多数月卡用户并非全天均在停车场内停车，因此停车场产生了数量较多的空闲车位。本文从这一背景出发，引入收益管理中的超售思想，以期将空闲车位加以利用，实现停车场收益的最大化。本文基于历史停车数据并结合收益管理思想，区别于以停车位利用率最高为目标的传统停车场运营管理研究，以停车场收益最大化为目标对停车场月卡超售问题进行优化。此外，本文考虑了停车场销售月卡以及临时停车两种收益来源以及与之相对应的两类停车用户，更加贴合运营现状。结合停车场运营公司的实际收费策略，本文通过0-1变量表示在场与不在场两种状态，将停车时长转化为时间序列，并对时间序列相同的数据赋予相同的标签，定义了停车模式。进而使用KNN和K-means算法对停车模式进行聚类，使停车模式能更好地反映停车用户的群体行为特征。根据停车场的实际运营情况，本文建立了基于停车模式的停车场月卡超售优化模型。数值实验结果显示，在停车需求相同的情况下，执行超售策略可以使停车场获得更高的收益。在此基础上，本文研究了停车需求不确定性的场景。考虑到停车场的公共交通设施特性，本文建立了两阶段鲁棒优化模型以使得在超售策略下停车场运营受到停车需求不确定性的影响最小。在两阶段问题中，第一阶段决策月卡销售数量，第二阶段用违约的惩罚对月卡销量进行约束。由于两阶段鲁棒优化模型中含有整数决策变量，因此本文使用了嵌套列与约束生成算法对模型进行求解。实验结果表明，本文所提出的模型能够在需求不确定的情况下获得较优的收益。此外，基于停车模式的优化模型在较短的时间内即可处理实际规模的问题，其所具有的高效性使得该模型可以较好地应用于实际场景之中。

With the rapid development of Chinese economy and the overall progress of the automobile industry, the number of motor vehicles in China has increased year by year. But what follows is a serious shortage of parking spaces in China, and the supply and demand of parking facilities are seriously unbalanced. At the same time, with the rapid development of my Chinese parking market, more and more attention has been paid to the revenue management of parking lots. In the operation of parking lots in urban core areas, the customer base is relatively fixed, so the parking lot has launched monthly parking card products to meet customer needs. But most monthly card users do not park in the parking lot all day, so the parking lot produces a large number of vacant parking spaces. Starting from this background, this paper introduces the idea of overbooking in revenue management, in order to maximize the revenue of the parking lot by utilizing the vacant parking spaces.Based on historical parking data and combined with revenue management ideas, this paper optimizes the parking lot monthly card overbooking problem with the goal of maximizing the parking lot revenue, which is different from the traditional parking lot operation management research that aims at the highest utilization rate of parking spaces. In addition, this paper considers the two revenue sources of monthly card sales and temporary parking in the parking lot and the corresponding two types of parking users, which is more in line with the current operating situation.Combined with the actual charging strategy of the parking lot operating company, this paper uses 0-1 variables to represent the two states of presence and absence, con-verts the parking duration into a time series, and assigns the same label to the data with the same time series, defining the parking pattern. Furthermore, the KNN and K-means algorithms are used to cluster the parking patterns, so that the parking patterns can better reflect the group behavior characteristics of parking users.Based on the actual operation of the parking lot, this paper establishes an overbook-ing optimization model which based on the parking pattern. The numerical experiment results show that implementing the overbooking strategy under the same parking de-mand can make the parking lot get a higher income. On this basis, this paper studies the scenario of parking demand uncertainty. Considering the characteristics of public trans-portation facilities in the parking lot, this paper establishes a two-stage robust optimiza-tion model to minimize the impact of parking demand uncertainty on parking lot opera-tions under overbooking strategies. In the two-stage problem, the first stage determines the monthly card sales volume, and the second stage uses the penalty for breach of con-tract to constrain the sales. Since the two-stage robust optimization model contains inte-ger decision variables, this paper uses nested columns and constraint generation algo-rithms to solve the model. The experimental results show that the two-stage robust op-timization model can obtain better returns under the condition of uncertain demand. In addition, the optimization model based on parking patterns can handle real-scale prob-lems in a short time, and its high efficiency makes the model suitable for practical sce-narios.