Risk-driven smart buffer optimization for metro crowd control

The collaborative management of diversion facilities in passenger flow buffer zones has become an important response strategy for high-density crowd risk prevention in metro stations. To address this, a partition independent principal component analysis (PIPCA) method for control schemes in the buffer zone scenario of metro stations is proposed, and a new high-density crowd risk assessment model is designed by using the moss growth optimization (MGO) to optimize the NARX model. Furthermore, the value function in the model predictive control is improved to adaptively optimize the flow lines in the buffer zone. A simulation experiment is conducted on a Qingdao metro station. The results show that: (1) the proposed PIPCA-MGO-NARX model performs well in prediction performance with significantly reduced MSE, demonstrating excellent prediction accuracy. (2) The constructed risk assessment model identifies crowd risks under different buffer zone schemes through quantitative analysis, providing a basis for the implementation of management strategies. (3) The proposed diversion facility management strategy reduces the frequency of facility changes in different zones by 33.3%–50% compared to the MPC method, while ensuring traffic efficiency and safety in advance, demonstrating better practicality. This research provides important solutions for addressing the challenges of high-density crowd safety control.