Parameter sensitivity analysis of a reduced-order electrochemical-thermal model for heat generation rate of lithium-ion batteries

A reduced-order electrochemical-thermal model is capable of accurately estimating heat generation rate of lithium-ion batteries with high accuracy. However, as the number of parameters drastically increases, an identification of the parameters becomes essential but challenging. In this paper, a sensitivity analysis of its physical parameters is conducted to study the influence of the parameters on the heat generation rate for the first time. First, 23 potential parameters are selected, where the ranges of the parameters are assumed based on the literature. The parameters are then varied, and the corresponding heat generation rate is calculated to quantify their effects. The analysis has shown various tendencies of the sensitivity of the parameters, and their behavior is explained from further analysis. Based on their sensitivities, the parameters are clustered into three groups dependent upon dominant ranges of state of charge, where insensitive parameters are excluded. Finally, the parameters in each group are identified in their dominant ranges of the state of charge separately based on a three-stage identification procedure. For the identification, the genetic algorithm is used, where the objective function is to minimize the model error for the battery temperature. The results of the identification have shown that the parameters are converged within an optimal solution. The model with the identified parameters is further validated against the experimental data using a pouch type large format lithium-manganese-cobaltoxide/ Graphite cell at constant charge and discharge profiles over the full range of the state of charge and driving cycles, which has shown a good agreement.

Automated summary

The study conducted a sensitivity analysis of a reduced-order electrochemical-thermal model to investigate the influence of physical parameters on heat generation rate of lithium-ion batteries. Parameters were clustered into three groups and identified based on dominant ranges of state of charge, leading to convergence within an optimal solution. The model with identified parameters was validated against experimental data, showing good agreement.