当城市轨道交通系统出现异常，乘客容易产生盲目、恐慌、从众等心理，并做出非完全理性的决策。现 有MNL模型基于信息完备假设与完全理性假设，在面对非常态情景时的适应性较差。为刻画非常态情况下乘客 的不完全理性并考虑乘客个体差异，解决MNL模型的不适用性，本文针对轨道交通非常态情景，通过累积前景 理论进行建模。首先，综合考虑时间、费用、舒适度与便捷性4个对乘客出行行为有直接影响的因素，构建基于 累积前景理论的轨道交通非常态乘客出行方式选择模型，刻画乘客的不完全理性行为。然后，开展问卷调查用以 标定模型参数，并结合调查结果构建服从泊松分布的差异化参考点，描述模型的参考点依赖现象，差异化参考点 的泊松分布检验值满足大于等于0.05的检验标准，解释了乘客出现不同决策结果的本质，使综合前景呈现随参考 点波动的态势。最后，将引入差异化参考点的基于累积前景理论的轨道交通非常态乘客出行方式选择模型与多项 Logit(multinomial logit，MNL)模型进行对比。研究结果表明：针对非常态情景，本文模型体现了乘客面对非常态 时的有限理性与个体差异性，宏观角度下的交通方式划分率计算结果总体偏差较MNL模型减小了4.9%，微观角 度下乘客个体决策结果预测准确率较MNL模型提升了25.4%，整体效果优于MNL模型，验证了模型的合理性。 研究成果可以为轨道交通非常态下的交通需求预测与应急响应提供理论支撑。

When a traffic system is abnormal, passengers are prone to blindness, panic, conformity, and other psychological problems. They may thus make incomplete rational decisions. The Multinominal Logit (MNL) model is based on the assumptions of complete information and rationality. It has poor adaptability when used for abnormal situations. Therefore, the incomplete rationality of passengers under abnormal conditions was described using the cumulative prospect theory, and individual differences among passengers were considered to resolve this inability of the classical MNL model. First, the four factors of time, cost, comfort, and convenience were comprehensively considered, and a model was constructed of rail transit non-normal passenger travel mode selection based on the cumulative prospect theory. It was used to characterize the incomplete rationality of passengers. Afterward, a questionnaire survey was conducted to calibrate the model parameters. Based on the survey results, a differentiated reference point following a Poisson distribution was obtained to describe the reference point dependency phenomenon of the model. The results of a case study indicated that the Poisson distribution test values with the introduction of differentiated reference points met the test criterion of a value that was greater than or equal to 0.05. It explained the essence of passengers’ different decision-making results and presented a trend of comprehensive prospects fluctuating with the reference points. Finally, this model was compared with the MNL model to verify the rationality of the model. The research results indicated that the model focused on abnormal situations and reflected the incomplete rationality and individual differences of passengers. The overall accuracy was higher than that of the MNL model, and the average absolute error was reduced by 4.9%. The accuracy of the microscopic calculation results was 25.4% better than that of the MNL model. This could provide theoretical support for traffic demand prediction under abnormal rail transit conditions.

U491

国家重点研发计划(2019YFF0301403)，中央高校基本科研业务费专项资金 (2019JBM041)