Robustness enhanced capacity estimation method for lithium-ion batteries based on multi-voltage-interval incremental capacity peaks

Abstract

Accurate battery capacity estimation can contribute to safe and reliable operations of lithium-ion battery systems. The incremental capacity (IC) based techniques provide promising estimates of battery capacity. However, curve smoothing algorithms are usually required in the IC-based methods, which introduce additional errors and are computationally burdensome. To address this issue, this work proposes a novel approach using multi-voltage-interval IC peaks combined with a back-propagation neural network (BPNN) for battery capacity estimation. Multiple voltage intervals covering relatively narrow and wide values are applied for computing IC curves to enhance the estimation robustness. In particular, there is no need to employ smoothing algorithms. A BPNN is then applied to approximate the correlation between multi-voltage-interval IC peak and capacity. Besides, a five-point moving window technique is proposed to capture multi-voltage-interval IC peaks online effectively. Experimental results show capacity estimates with the majority of relative errors of ±1% and the maximum error of 2%.