A novel lithium-ion battery state of charge estimation method based on the fusion of neural network and equivalent circuit models

Accurate State of Charge (SOC) estimation plays a crucial role in ensuring the safety and efficiency of lithium-ion batteries. The existing data-driven SOC estimation methods usually use the directly measured feature information to estimate SOC, resulting in a large calculation error. This paper presents a SOC estimation method which combines the equivalent circuit model (ECM) with the gated cycle unit (GRU) neural network. Firstly, the terminal voltage of the battery is obtained independently through a second-order resistance-capacitance model and GRU. Then, the error weight method is introduced to process the voltage data obtained by ECM and GRU, which help to reduce the influence of terminal voltage error on the final SOC estimation result. To improve the robustness of the system, the difference between the processed terminal voltage and the actual output voltage is synchronously imported into the GRU as an additional feature to increase the input dimension of the model data. Lastly, the proposed SOC estimation method is evaluated under the dynamic current condition. Compared with the single data-driven method, the root-mean-square error of the proposed SOC estimation method is reduced by 1%. The results show that the proposed method has better accuracy.