Cross-Term Correlations Considered Extended Kalman Filter for High-Speed Train State Estimation

Accurately assessing the operational state of high-speed trains (HST) is crucial for passenger safety and operational efficiency. This paper develops a model that integrates the practical aspects of train operation and introduces a higher-order extended Kalman filter that accounts for cross-term correlations. By effectively incorporating higher-order terms, which are typically neglected in conventional extended Kalman filters, and considering the potential correlations between higher-order terms and noise components, the method enables the comprehensive utilization of information that is often overlooked by other higher-order approaches. Moreover, given that premature transitions between the train’s operational modes can lead to significant inaccuracies in state estimation, this paper proposes a sliding window error model designed to adapt to the train’s operational modes in real-time. The simulation results validate the effectiveness of this methodology in estimating train states. Additionally, to ensure applicability in real-world scenarios, the study uses operational data from the CRH2C train running on the railway from Zhengzhou to Xi’an as a simulation object, confirming that the proposed method remains effective in real situations. Furthermore, even in the presence of anomalies in the monitoring data, the proposed technique maintains state estimation accuracy that meets the requirements of the actual system, thereby affirming the method’s effectiveness.