(457c) Robust Surrogate Modeling of Lithium Ion Batteries

Lithium ion batteries hold high promises for the next generation of hybrid vehicles such as fuel cell/battery hybrid systems. Hence, understanding these energy storage devices is of paramount importance. Physics-based models available in literature for lithium ion batteries have high fidelity, but they are computationally intensive and also are difficult to integrate into hybrid vehicle system models. Moreover, these physics-based models are deterministic and uncertainty analyses become difficult. Hence robust designs with metamodels or surrogate models are the need of the hour. One of the most prominent surrogate modeling techniques is the Response Surface Methodology (RSM). RSM is a collection of statistical and mathematical techniques useful for developing, improving and optimizing processes [1]. Surrogate models also help in reducing the curse of dimensionality. These models will help us evaluate the influence of different power sharing algorithms and control architectures in a hybrid system on the battery performance earlier in the design phase efficiently and with accuracy.

In this work, a Central Composite Design of experiments is chosen to obtain discharge performance curves of LiFePO4 batteries as a function of the parameters-temperature, discharge rate and depth of discharge. The responses that are of interest are the discharge capacity, specific power and specific energy. Typical performance curves are shown in Figure 1.  The RSMs that are developed will further help us understand the relative influence of each of the parameters on the responses.

Figure1. Performance characteristics of

3Ah LiFePO₄ cell.

Reference:

[1]Raymond H. Myers,
Douglas. C. Montgomery, Response Surface Methodology, Second Ed., Wiley,
New York, 2002, pp. 1.

Acknowledgements:

Georgia Tech Research Institute is acknowledged for funding of this work.