Real-Time Deployment Strategies for State of Power Estimation Algorithms

Lithium-ion cells operate under a narrow range of voltage, current, and temperature limits, which requires a battery management system (BMS) to sense, control, and balance the battery pack. The state of power (SOP) estimation is a fundamental algorithm of the BMS. It operates as a dynamic safety limit, preventing rapid ageing and optimizing power delivery. SOP estimation relies on predictive algorithms to determine charge and discharge power limits sustainable within a specified time frame, ensuring the cell design constraints are not violated. This paper explores various approaches for real-time deployment of SOP estimation algorithms for a high-power lithium-ion battery (LIB) with a low-cost microcontroller. The algorithms are based on a root-finding approach and a first-order equivalent circuit model (ECM) of the battery. This paper assesses the practical application of the algorithm with a focus on processor execution time, flash memory and RAM allocation using a processor-in-the-loop (PIL) setup. The case study estimates the maximum power available for regenerative braking at high SOCs and compares predictions with experimental data. More specifically, deployments using single and double-precision floating numbers are compared, alongside different voltage estimation approaches. In addition, the bisection root-finding method is compared to the secant and Brent’s method. The different algorithms tested in this study do not significantly impact memory allocation. In terms of processor load, however, single-precision deployments are significantly more cost-effective than double-precision deployments, with a negligible discrepancy in the predicted output. Finally, the secant root-finding method reduces the execution time by two-thirds while retaining the same level of accuracy when compared to the bisection method.