Methodology and Guidelines for Designing Flexible BMS in Automotive Applications

Abstract

The fragile characteristics of Li-ion batteries lead to the need of battery management system (BMS) to carefully supervise them during the operation. Since there are so many variations in battery configurations, the BMS usually must undergo many iterations of the development cycle, which take a long time to optimize and finalize the design. Previously, many works adopted the idea of modularized BMS to address these issues, but they still have some skeptical issues such as measurement approaches or difficulties in reconfiguration. This paper presents a guideline on the crucial aspects of flexible BMS designs for automotive applications, which aims to reduce time and effort for developing a new BMS for automotive battery pack. The guideline covers some crucial aspects pertaining the automotive BMS hardware implementation, SOC estimation algorithm and its computational performance based on Extended Kalman Filter (EKF) and Luenberger Observer (LO) with 3 levels of Electrochemical model (ECM). All of the tests were carried out in a small-scale microcontroller. It was found that 2-RC ECM gives the best trade-off between SOC estimation accuracy and computational time. While the 3-RC ECM provides 9.5% and 31% higher accuracy than the 2-RC and 1-RC ECM, respectively, but taking 88% and 240% higher computational time than the latter two cases. The optimal speed of the observer poles of LO algorithm are suggested to be in the range of 2-5 times faster than the system poles, which makes the convergence speed to be comparable to the EKF algorithm but is still able to keep the SOC estimation error in the range of 3-5%. These results can be used to make a trade-off between estimation accuracy and computational time, to select the optimal SOC estimation algorithm for onboard BMSs.