Parameter Identification for Electric-thermal Model of Lithium Ion Batteries


Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
Award date2 Jun 2017


Lithium-ion batteries (LIBs) have been widely utilized as power sources of electrical vehicles (EVs) and hybrid electrical vehicles (HEVs) in recent years, due to their advantages in higher energy-to-weight ratios, longer life cycle and lower environmental pollution. The increasing demand has fueled the need for the improved safety and performance of LIBs, which are all related to the thermal performance. A battery thermal management system that keeps the battery to work within an optimal temperature range is crucial in EV/HEV application.

The battery electric/thermal process is a typical distributed parameter system (DPS) which is spatiotemporal coupled. An accurate model to simulate the electric/thermal process of LIBs requires knowing a great number of physical properties, which are difficult to obtain directly due to the following challenges and difficulties:

(1) The battery electric/thermal model is extremely complicated with high-dimensional nonlinearity, the multi-scale (global and local) complexity exists in the parameter identification.

(2) Due to the high correlation between parameters, and the large search hyperspace, the identification is very complicated if all the parameters are identified synchronously.

(3) The electro-chemical behaviors in batteries are very sensitive to the external disturbances and variation of operational conditions, the developed model should be updated to detect the dynamic environment.

(4) For the thermal model, the numerical solution needs exhausted computation, an efficient method needs to be proposed to obtain the analytical solution for the online identification.

(5) When the battery is charging or discharging, the electric behavior and thermal behavior influence each other. To obtain a more accurate identification result, a method need to come out for this electric-thermal coupled multi-physics model.

Despite of the discussed difficulties, there is still an open area for estimation of the electric/thermal behavior in lithium-ion battery systems. The main objective of this research is to develop an efficient parameter identification algorithm and predict the accurate discharging behaviors for the lithium ion batteries. The estimated distributions can be used for control, health monitoring and design purposes. The proposed method need to handle these difficulties and obtain an accurate predictions of model output:

(1) Though the battery electric model is extremely complicated with high-dimensional nonlinearity, it can still be approximately linearized in a neighborhood of any given operating point based on the theory of linear system. Thus, the identification can be considered as a multi-scale problem (global and local). A weighted PSO-LM algorithm for the multi-scale (coarse and fine) identification of parameter vector x by using a combination of the particle swarm optimization (PSO) algorithm and the Levenberg-Marquardt (LM) algorithm is proposed. PSO is firstly utilized to implement the coarse-scale parameter identification to find the near-optimal point in the global space, and then the LM algorithm is embedded in PSO to conduct the fine-scale parameter identification in the vicinity of the optimum within the local space.

(2) To address the overwhelming complexity of identifying parameters simultaneously, the number of identified parameters can be reduced during each identification. A global sensitivity analysis (GSA) prior to the parameter identification provides valuable insights into the identifiability of the model parameters. The parameters are then sorted and grouped according to the sensitivity indices. The LM algorithm is applied to optimize the parameters group by group. Numerical simulation results demonstrate that the group-wise identification algorithm can serve as a reliable tool for extracting parameters.

(3) The identified offline model should be updated online to adapt to the external disturbances and variation of working environments. The proposed clustering PSO algorithm not only finds the global optimal solution under a specific environment but also tracks the trajectory of the changing optima over dynamic environments. The effectiveness of the online identification method has been validated with experiments. The developed method is simple and can be adaptive for online application.

(4) A simple but effective nominal model is first developed for real-time parameter estimation of thermal model using the Karhunen-Loeve decomposition, which is applied to simplify the thermal model. The LM based online parameter identification is performed to estimate the thermal parameters. The simulations results demonstrate the effectiveness of the proposed method.

(5) To analyze the electric/thermal performance simultaneously, a multi-objective parameter optimization method is presented to model the battery performance by fitting the simulated terminal voltage and surface temperature with the measured value simultaneously. The pareto-front is obtained by the multi-objective genetic algorithm NSGA-II to select the optimal individual to fit both voltage and temperature well.