Comprehensive Design of an Electric Vehicle Battery Pack and Management System with Microprocessor Memory Use Analysis
This paper investigates the memory usage of battery management system (BMS) software functions required for operating the high-voltage battery system (HVBS) of a light-duty electric vehicle (EV). Seven BMS software functions—battery balancing and monitoring, state of charge (SoC) and power estimation, fault diagnosis, contactor control, and communication management—are developed, validated through processor-in-the-loop (PIL) testing, and evaluated for memory usage when deployed on the automotive-grade NXP S32K358 microprocessor unit. A comprehensive high-voltage battery and management system for the RAM ProMaster EV is developed to ensure the memory assessment reflects real-world deployment conditions, including a detailed mechanical and electrical design of the battery module, battery disconnect unit (BDU), high-voltage mechanical structure, and a thorough software architecture. A high-fidelity plant model of the HVBS is implemented in the MATLAB/Simulink 2023b environment to capture the electrical and structural characteristics of the system, enabling validation of the BMS software functions prior to deployment. An experimental dataset for 21,700 cylindrical lithium-ion cells from the RAM ProMaster EV is used to parameterize and test state estimation algorithms. The dataset covers a wide range of operational conditions, including multiple standard and non-standard drive cycles (UDDS, LA92, US06, HWFET, HWCUST and HWGRADE) and characterization tests (C/20 discharge, four-pulse HPPC and GITT). The results show that the SoC estimation system accounts for the majority of the memory used by BMS software functions, reaching 31.4 kB of flash and 52.2 kB of RAM, while the remaining functions exhibit significantly lower memory demands, consuming no more than a half of that amount. Overall, the findings suggest that the BMS software functions occupy a relatively small portion of the S32K358’s 8 MB of flash and 1125 kB of RAM (0.84%-flash and 8.63%-RAM) compared with the system overhead (3.89%-flash and 6.30%-RAM), leaving sufficient margin for integrating more advanced estimation and protection functions in future EV battery systems.