Achieving precise prediction of short-term inbound and outbound passenger flows in urban rail transit is of great significance for improving train transportation efficiency and enhancing passenger travel quality. This paper aims to analyze the influencing factors of short-term passenger flows and utilize Spearman correlation analysis to screen characteristic variables. Based on this, a short-term inbound passenger flow prediction model for urban rail transit is constructed, combining an improved Particle Swarm Optimization (PSO) algorithm with a Long Short-Term Memory (LSTM) network. Furthermore, to address complex and ever-changing scenarios, this paper introduces a digital twin perception function that can dynamically optimize the key parameters of the prediction model, ensuring its robustness and adaptability, and improving the accuracy of prediction results. Experimental results demonstrate that, compared to traditional prediction methods, the model proposed in this paper exhibits superior performance in terms of average absolute error and goodness-of-fit value, showcasing stronger generalization capabilities and adaptability to various conditions.