Microprocessor Execution Time and Memory Use for Battery State of Charge Estimation Algorithms 2022-01-0697

Accurate battery state of charge (SOC) estimation is essential for safe and reliable performance of electric vehicles (EVs). Lithium-ion batteries, commonly used for EV applications, have strong time-varying and non-linear behaviour, making SOC estimation challenging. In this paper, a processor in the loop (PIL) platform is used to assess the execution time and memory use of different SOC estimation algorithms. Four different SOC estimation algorithms are presented and benchmarked, including an extended Kalman filter (EKF), EKF with recursive least squares filter (EKF-RLS) feedforward neural network (FNN), and a recurrent neural network with long short-term memory (LSTM). The algorithms are deployed to two different NXP S32Kx microprocessors and executed in real-time to assess the algorithms' computational load. The algorithms are benchmarked in terms of accuracy, execution time, flash memory, and random access memory (RAM) use. In order to ensure the validity of running these models for multiple cells in the pack, the impact of increasing the number of instances to run each algorithm simultaneously is investigated as well. The results show that the four algorithms present a reasonable accuracy, with less than 5% maximum error. For the more power microprocessor tested, the execution time was found to be 0.24 ms, 0.25 ms, 0.14 ms, and 0.71 ms for the EKF, EKF-RLS, FNN, and LSTM respectively. The neural network SOC estimation algorithms were also demonstrated to have lower RAM use than the EKFs, with less than 1 kB RAM required to run one instance of the estimators. Moreover, the FNN SOC estimation algorithm is found to be a promising option with both low execution time and memory use.