Multiscale Parameter Identification Algorithm with Dynamic-Tracking for Distributed Electric Model of Lithium-Ion Battery

摘要

This paper proposes a sentry particle based dynamic-tracking particle swarm optimization-Levenberg-Marquardt (PSO-LM) algorithm to identify the parameters for distributed electric model of Lithium-ion battery (LIB). In the proposed algorithm, a sentry particle is applied to track the dynamic changes of the working environment caused by observation noise or time variation of a system. A sliding window with suitable window size is applied to collect the experimental data for online estimation. Each time when a dynamic change is detected, the PSO-LM algorithm will be adopted to find the global optimal of parameters in the model. Specifically, PSO is firstly used for the coarse-scale parameter identification and LM is then conducted for the fine-scale parameter identification. The results demonstrate the effectiveness and efficiency of the proposed method. The multiscale identification algorithm with dynamic tracking herein can be also modified and further applied to various types of LIB models in driving circumstances.