Online state of charge estimation for lithium-ion batteries using improved fuzzy C-means sparrow backpropagation algorithm

Abstract

With the rapid development of new energy vehicles (EVs), cloud-based management of the lithium-ion batteries (LIBs) state of charge (SOC) has become the technological mainstream under increasing intelligence. However, SOC is highly sensitive to the modeling approach and data volume, making high-precision real-time estimation under complex conditions a significant challenge. An online estimation model based on an improved fuzzy C-means clustering sparrow search algorithm with a backpropagation neural network (FCMC-SSA-BP) has been developed to address this issue. The model collects raw voltage and current data through vehicle control speed in real-world conditions, which is then denoised and transmitted in real-time to an online cloud-based high-precision estimation system. The estimated remaining battery capacity is subsequently sent to the online battery management system (BMS) and visualized. The performance of the algorithm within the system directly influences the accuracy of SOC estimation. In the BP model, momentum factors and weight correction are introduced to enhance the stability of gradient learning to data volume. The efficiency of the algorithm is further improved using an enhanced Logistic chaotic map and an advanced elite reverse learning strategy. Additionally, the modified FCMC algorithm is employed to reduce the impact of nonlinear characteristics on prediction accuracy. Finally, the test results showed that the maximum error (Max_E) of the IFCMC-SSA-BP reached 0.84 %, 0.52 %, and 0.0031 % at 0 °C, 15 °C, and 35 °C under BBDST. Similarly, it reached 6.82 %, 3.29 %, and 1.4 % under HPPC, and for UDDS condition, it reached 9.33 %, 4.95 %, and 4.88 % at 20 °C, 25 °C, and 30 °C.