Digital twin of a battery for traction rolling stock and substations: remaining useful life estimation based on a dual Kalman filter

Background: One of the key objectives of digital transformation in railway transport is the development of digital platforms and digital twins aimed at improving operational efficiency. Predicting the remaining useful life of equipment is a critical task for traction rolling stock and traction substations of traction power supply systems. A key component affecting the reliability of these assets is batteries. The development of a digital twin of a battery for traction rolling stock and substations enables the estimation of essential parameters, namely remaining capacity and remaining useful life (RUL).

Aim: To develop a digital twin of a battery for traction rolling stock and traction substations of traction power supply systems, providing estimates of remaining capacity and RUL based on a dual Kalman filter, using a mixed approach to state-of-charge (SoC) estimation and adaptive forgetting factors derived from measurements of current, voltage, and temperature.

Materials and methods: To achieve the stated aim, the research employed methods of experimental design, mathematical statistics, differential calculus, and a dual Kalman filter algorithm. Verification of the results was performed using experimental data from a lithium-ion battery with a series connection of 15 cells. The estimation of remaining capacity and RUL was based on joint filtering of the parameters of a Thevenin equivalent circuit model and the SoC.

Results: A digital twin of a battery for traction rolling stock and substations has been developed. The proposed digital twin enables the determination of remaining capacity and RUL from measured values of current, voltage, and temperature.

Conclusion: The results were verified against measurement data from standard control and monitoring systems. The mean absolute percentage error for the SoC and voltage estimation was less than 1.5% and 0.5%, respectively. The obtained estimates of the RUL and remaining capacity are reliable and consistent with the results of similar studies in charge–discharge cycles. The proposed method for estimating the remaining useful life and remaining capacity of batteries for traction rolling stock and substations is recommended for use in technical condition monitoring and diagnostic systems within the framework of the Digital Locomotive and Digital Substation projects.