State-of-charge estimation of lithium-ion batteries based on gated recurrent neural network

Accurate state-of-charge (SOC) estimation, which is critical to ensure the safe and reliable operation of battery management systems in electric vehicles, is still a challenging task due to sophisticated battery dynamics and ever-changing ambient conditions. In contrast to model-based SOC estimation methods, whose performance rely heavily on the quality of battery models, neural network-based methods are purely data-driven and model-free, and can be easily extended. Recently, with the ever-increasing computing power provided by graphic processing units, the neural network-based methods have gained more and more attentions. In this paper, a recurrent neural network with gated recurrent unit is proposed to estimate the battery SOC from measured current, voltage, and temperature signals. Compared with traditional feed-forward neural networks, the proposed method exploits information of the previous SOCs and measurements and yields better estimation accuracy. The proposed method presents satisfying estimation results on data collected from two mainstream lithium-ion batteries under dynamic loading profiles. Moreover, the proposed method is robust against unknown initial SOC values and can be trained to learn the influence of ambient temperatures. The proposed method can estimate the SOC at varying temperatures with root mean square errors within 3.5% and works under untrained temperatures.