A firefly-based particle filter technique for system state estimation and battery RUL prediction

Abstract Accurate measurement and estimation of state variables in nonlinear dynamic systems are critical in various engineering and scientific applications. While particle filter (PF) techniques have become increasingly popular for modeling such dynamic systems, they are prone to sample degeneracy and impoverishment, which can considerably affect performance. Existing PF methods often come with drawbacks like high computing costs and noise sensitivity. This paper presents a novel probabilistic posterior optimization particle filter (PPO-PF) technique, inspired by the firefly algorithm, to address these PF limitations and enhance PF efficiency. In the proposed PPO-PF technique, an adaptive search method is proposed to locate the high-likelihood region within the posterior space for improved convergence. A new step size selection method is suggested to facilitate navigation within the search space. In addition, a particle position optimization approach is proposed to guide low-weight particles towards high-probability regions, optimizing the posterior probability density function and mitigating sample degeneracy. The proposed PPO-PF technique is validated by simulation under various model conditions, including predicting the remaining useful life of Lithium-ion (Li-ion) batteries. The results indicate that the proposed PPO-PF reliably captures system dynamics with improved accuracy, even under high noise conditions. These findings highlight its potential to enhance measurement science by offering a more reliable approach for state estimation in complex systems.