知识计算与数值优化融合的铁路网列流推算研究OA

Accurate estimation of railway network train operation and regional capacity utilization is essential for supporting railway transportation situation simulation,train organization optimization,and related tasks.This study integrated multi-source data,including train operation records,arrival and departure reports,loading and unloading reports,and physical network topology.Through data preprocessing and knowledge fusion,freight path knowledge triples were extracted to construct a knowledge graph.By leveraging the knowledge graph for knowledge extraction and inference,historical train operation information and temporal parameters were obtained to build a freight service network for path searching,while historical freight car trajectory knowledge was additionally incorporated to supplement the candidate path set.Based on the parameters and path sets derived from knowledge computation,a linear integer numerical optimization model was constructed with the objective of minimizing total train travel time.Meanwhile,residual train flows were calculated through time judgment and path-cutting operations to realize continuous time-period train flow estimation.Finally,case studies were constructed using real production data,and the model was solved and verified using the commercial solver Gurobi.The results show that,compared with actual data,the estimation error of more than 85%of arcs is less than 5 trains,and the proportion of arcs with errors exceeding 10 trains is less than 2.5%.The bottleneck sections and lines identified by the estimation are generally consistent with actual conditions,providing a reference for train organization optimization.Compared with the estimation method based on the strategy of prioritizing the allocation of train groups with the largest number of pending freight cars,the proposed method improves accuracy by approximately 12%and reduces the number of arcs with large errors by 56.3%.Compared with a single numerical optimization method,the proposed method achieves an accuracy improvement of approximately 10%with a smaller candidate path set,and reduces the number of arcs with large errors by 58.2%.Furthermore,compared with the two benchmark methods,the paths estimated by the proposed method are more reasonable,and the resulting transportation schemes are more consistent with actual operations.An analysis of error distribution and causes indicates that factors such as data quality and optimization algorithms may affect estimation results,providing directions for further improvement.The research results provide decision support for railway transportation situation simulation,bottleneck alleviation,and train dispatching,and contribute to the optimization of transportation organization schemes.