Date of Award

1-1-2019

Publication Type

Master Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

First Advisor

Balakumar Balasingam

Keywords

Expectation Maximization, Kalman filter, Li-ion, Online battery capacity estimation, Online open circuit voltage characterization, State of Charge Tracking

Rights

info:eu-repo/semantics/embargoedAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

One of the key steps of any battery management system design is the representation of the open circuit voltage (OCV) as a function of the state of charge (SOC). The OCV-SOC relationship is very non-linear that is often represented using a polynomial that has log and inverse terms that are not defined around SOC equal to zero or one. The traditional response to this problem was only at the software level. In this thesis, I present a formal scaling approach to the OCV-SOC characterization in Li-ion batteries. I show that, through formal modeling and optimization, the traditional approach to OCV-SOC modeling can be significantly improved by selecting the proper value of $\epsilon$. When the proposed technique is used a decrease in the maximum SOC error of 9\% is reported. The proposed approach is tested on data collected from multiple cells over various temperatures for OCV-SOC characterization and the results are presented. State-space model (SSM) and the Kalman filter have several applications in the emerging areas of automation and data science including in battery SOC estimation. In many such applications, the application of Kalman filtering requires model identification with the help of the observed data. I present the formulas with derivations for linear state-space model parameter estimation using the expectation maximization (EM) algorithm. Particularly, I derive the formulas for different special SSM cases of practical interest, such as the continuous white noise acceleration (CWNA) model. Through simulation, I show the benefits of these derivations for the special models in comparison with the generalized approach.

Available for download on Monday, January 18, 2021

Share

COinS