Model Parameter Identification for Lithium Batteries Using the Coevolutionary Particle Swarm Optimization Method

Abstract

This paper uses the coevolutionary particle swarm optimization (CPSO) method to identify battery parameters. A parameter identification window (PIW), which has the features of a fixed data length and real-time response, is used to store a piece of data that indicates the battery operation at the current moment. CPSO uses the data in the PIW to dynamically identify the battery parameters. Each equivalent circuit model (ECM) parameter uses a separate parameter particle swarm (PPS) to optimize their values. In every algorithm cycle, each particle in every PPS only evolves one step. The currently evolved PPS uses the current optimal values of the other PPS in CPSO to evaluate all of the particles and to find the best particle. Every PPS is scheduled by the CPSO, dynamically evolves one by one, and converges in real time to its optimal value, which is an ECM parameter. Real battery data are used to test the algorithm. The experimental results indicate that the fluctuation patterns of the open circuit voltage (OCV) are accurately identified. For the different algorithm parameters, the identification results for the OCV have good consistency, and the deviations between the identification results are less than 5 mV most of the time.