A noise‐immune model identification method for lithium‐ion battery using two‐swarm cooperative particle swarm optimization algorithm based on adaptive dynamic sliding window

Abstract

Accurate and reliable model parameters are not only a prerequisite for model-based estimation but also a significant part of battery operating characteristics. However, the measurement signal inevitably contains noise, which brings great challenges to model identification. This paper focuses on the noise immunity performance of model identification based on two-swarm cooperative particle swarm optimization. An adaptive dynamic sliding window based on the current rate criterion and the identification results feedback is designed to avoid data redundancy and improve the robustness of model identification. The model parameters are obtained using two-swarm cooperative particle swarm optimization based on the adaptive dynamic sliding window. The proposed method effectively improves the accuracy and speed of parameter identification through optimization of data fragments and particle update rules. Compared with two existing parameter identification methods, simulation studies illustrate that the average mean square deviation of the proposed method is reduced by at least 35 dB. The proposed method is superior to existing parameter identification methods in noise immunity performance, parameter identification reliability, and state-of-charge estimation accuracy. By employing the proposed method, the maximum errors of state-of-charge estimation are limited within 1% under experimental verification. The experiment results verify that the proposed method has the potential to extract reliable model features online.