Abstract: In response to the bottleneck of the clearing, reconciliation and discrepancy handling services of the railway electronic payment platform under the centralized architecture, this paper proposes an application upgrade solution based on distributed parallel processing technology. A dual-data-center active-active deployment architecture is adopted, distributed data storage is achieved through data sharding and hash algorithms, and the processing time of batch services is effectively shortened by combining with the Spark stream-batch integrated computing engine; relying on the hot-cold data hierarchical storage mode to achieve cost reduction and efficiency improvement, a full-process intelligent reconciliation system integrating multi-source data collection, standardized data conversion, configurable reconciliation engine, intelligent discrepancy classification and automatic discrepancy adjustment is built to improve the automation level of reconciliation and the accuracy of accounting discrepancy identification. Meanwhile, a full-link visual monitoring platform is built to realize pre-warning and active prevention and control of business risks. Practical application proves that this solution can stably support the high-concurrency access demand during business peaks, significantly enhance the stability of system operation, reduce the occurrence rate of accounting discrepancies, and comprehensively protect the security of railway transportation revenue funds.