A SOC Estimation Method for Li-Ion Batteries under High-Rate Pulse Conditions based on AO-BPNN Model

The estimation of state of charge (SOC) in lithium-ion batteries is important for ensuring the safe and stable operation of battery systems. Under high-rate pulse conditions, the characteristics of short discharge time, high frequency, large current, strong interference, and complex transient characteristics that make lithium-ion batteries exhibit marked nonlinear characteristics. The existing battery management system has difficulties in capturing the rising and falling edge data of the pulses due to limitations in the sampling frequency. The short idle time makes it challenging to obtain accurate open-circuit voltage, and there are difficulties in identifying the model parameters. Therefore, using a combination of coulomb counting method, open-circuit voltage correction method, and Kalman filtering method to estimate SOC poses certain challenges. This study applies backpropagation neural network (BPNN) combined with Aquila optimizer (AO) algorithm to estimate SOC under high-rate pulse conditions, and experimental verification is performed using special 3-Ah lithium iron phosphate battery. We compared the estimation accuracy of the AO-BPNN model for SOC with the BPNN, support vector machine, extreme learning machine, and Fuzzy neural network models and verified the superiority of AO-BPNN. Furthermore, by utilizing data with larger acquisition intervals, we obtained accurate evaluation results and reduced the data requirements. The effectiveness of the assessment of AO-BPNN was individually verified under different high-rate pulse conditions and different static times through pulse experiments conducted under 9 operating conditions, with the estimation error controlled within 5%. Finally, the robustness of the proposed model was validated using test data with different sampling intervals and random measurement errors.