A Comparative Investigation on Peak Current Solution Methods for Lithium-Ion Battery Peak Power Capability Prediction

Abstract

Accurate battery peak power capability prediction plays an essential role in improving the safety and efficiency of battery operations. The end of battery charge or discharge is caused by depleted or saturated surface lithium-ion concentrations of electrode solid particles to avoid damaging side reactions. Precise battery peak power capability prediction necessitates a high-fidelity electrochemical model (EM) that accurately depicts dynamic changes of lithium-ion concentrations inside a battery. One of the critical challenges to apply battery EMs for peak power prediction is how to accurately solve the peak charge and discharge currents from a set of complex model equations. To address the issue, this paper mainly investigates four different peak current solution algorithms, including bisection method, genetic algorithm method, particle swarm optimization method, and grey wolf optimizer (GWO) method for battery EM-based peak power prediction. The dependences of the prediction results using different current solution methods on the predictive time horizon and ambient temperature are validated. Their performance in terms of prediction accuracy, computational speed, complexity, convergence rate, and random-access memory are also evaluated. Among these peak current solution methods, the GWO method is proven to have a better performance in convergence rate and computing time while maintaining high prediction accuracy.