State of Charge Estimation of Lithium-Ion Battery Based on Multi-Modal Information Interaction and Fusion

The assessment of the charge status of lithium-ion batteries holds significant importance in the functioning of electric vehicles, particularly when it comes to real-time monitoring and ensuring safety control. In this regard, a backpropagation neural network based on multi-modal information interaction and fusion (MMI-BPNN) has been proposed, which incorporates the interaction and fusion of multi-modal information, to estimate the state of charge (SOC) of lithium-ion batteries. Initially, the principles of Unscented Kalman Filter (UKF) and backpropagation neural network (BPNN) are introduced. Building on this foundation, a specific approach for the interaction and fusion of multi-modal information in estimating battery SOC is presented. Then, different experiments are carried out under FUDS conditions, including UKF to estimate battery SOC, using first-order RC equivalent circuit and forgetting factor recursive least square method (FFRLS); BPNN estimation of battery SOC, using the data under the working condition for training; And the experiment involving the utilization of the MMI-BPNN technique for the estimation of battery SOC is conducted, employing information extracted from both the UKF and BPNN models. Finally, the error analysis conducted on the experimental results demonstrates that the maximum relative error in SOC estimation using the MMI-BPNN method is below 0.9325%. This finding confirms that the method exhibits superior accuracy and robustness when compared to the other two methods.