Fault detection and compensation capacitor capacity estimation in JTC based on parameter-optimized variational mode decomposition

To satisfy the safety development requirements outlined in the 14th Five-Year Plan, this study addressed the difficulties in fault identification and monitoring of the working conditions of compensation capacitors in jointless track circuits. Leveraging electromagnetic field theory and transmission line theory, a joint simulation model simulating track inspection vehicle technology was established to analyze the working states of compensation capacitors and perform fault identification and localization. First, to solve the issue of pre-setting the number of modes and penalty factors in Variational Mode Decomposition (VMD), a fitness function based on the “correlation center frequency ratio” was proposed for subsequent optimization calculations. Second, the dung beetle optimization algorithm was improved, and applied to optimize the VMD parameters using the proposed fitness function. Finally, VMD was performed on the simulation data with the optimized parameters, and features were extracted based on energy characteristics. These features represented the working states of the compensation capacitors, resulting in curves depicting the relationship between each capacitor ’ s characteristics and capacitance value. A capacitance estimation formula was derived from these curves and validated using real data from actual track circuits. The results indicate that the improved algorithm converges faster and achieves better optimization results as compared to the original and other mainstream optimization algorithms. The absolute error between the estimated and actual capacitance values was only 2.7 μF, and the fault identification accuracy reached 96.6%. The results indicate that the proposed method provides a reliable reference for fault identification and capacitance estimation of compensation capacitors in field conditions using inspection vehicle technology. © 2025, Central South University Press. All rights reserved.