A particle filter-based virtual sensor for estimating the state of charge and internal temperature of lithium-ion batteries: Implementation in a simulated study case

Battery management system (BMS) is a crucial tool to ensure correct operation and to prolong the useful life of lithium-ion batteries. Its main functionalities and decisions are based on accurate information provided by sensors of different types installed in the battery. Anomalies in the quantitative values of the main operational parameters of the cells can also be an indication of risk situations or imminent failures. Virtual sensors have become part of the BMS due to their advantages such as the absence of installation and maintenance processes and cost reduction with physical components. This paper proposes a Particle Filter (PF) based strategy to monitor simultaneously the state of charge and the internal temperature of a lithium-ion battery. The algorithm is chosen due to its ease of implementation, capacity to deal with non-linear and stochastic dynamics. Computational simulations are done in the software StarCCM+ with a battery cell operating with different C-rates to generate data to test the algorithm and to compare it with the consolidated Extend Kalman Filter (EKF) in terms of convergence velocity, error and computational cost. Similarly to the equation of the tested observers, the mathematical models of the electrical and thermal dynamics are presented as well as the Recursive Least Squares algorithm for the parameters identification. The observers were tested and both fulfilled the function of tracking the real-time values of state of charge and internal temperature. The final results show quantitatively that the PF algorithm presents a favorable convergence rate and accuracy when compared to the EKF for different current rates; however, it is necessary to control the number of particles used to find a processing cost closer to the EKF.

Automated summary

Battery management system (BMS) is crucial for the correct operation and prolonged life of lithium-ion batteries. It relies on sensors to provide accurate information and detect anomalies in cell parameters. This paper proposes a Particle Filter (PF) algorithm to simultaneously monitor the state of charge and internal temperature of a lithium-ion battery, showing favorable convergence and accuracy compared to the Extend Kalman Filter (EKF). Computational simulations were conducted to test the algorithm and compare it with the EKF in terms of convergence velocity, error, and computational cost.